Kurdistan Regional Government Ministry of Higher Education and Scientific Research University of Sulaimani College of Commerce
High Efficiency Watermarking Video Coding
A thesis Submitted to the Council of the College of Commerce, University of Sulaimani, in Partial Fulfillment of the Requirements for the Degree of Master of Science in Information Technology
By
Nyaz Aziz Ali
Supervised By Assistant Professor
Dr. Aree Ali Mohammed
September 2016
Rezber 2716
High Efficiency Watermarking Video Coding
A Thesis
Submitted to the Council of the College of Commerce, University of Sulaimani, in Partial Fulfillment of the Requirements for the Degree of the Master of Science in Information Technology
By
Nyaz Aziz Ali
Supervised By Assistant Professor
Dr. Aree Ali Mohammed PhD in Multimedia System
September 2016
Rezber 2716
سنَ مَا َعِملُوا ﴿ ﴾37لَِيجِسَِيهُمُ اللَّهُ أَحِ َ
ضِلهِ وَا َّللهُ َي ِرزُقُ َمنِ يَشَاءُ ِبغَيِرِ وَيسِيدَهُمِ ِمنِ فَ ِ حِسَابٍ ﴿﴾38
سورة النور
Acknowledgments First of all praise be to Allah for all his Graces in granting patience and faith to complete my present study.
I would like to express my special thanks and appreciation to my supervisor, Dr. Aree A. Mohammed for his guidance, encouragement, and support from the initial to the final level of this study, which enabled me to develop an understanding of the subject.
My Great thanks to Dr. Rezan HamaRashid, Mr. Haval M. Sidqi, Mr. Faraidoon Hassan and Mr. Delman Abdullah.
I would like to express my thanks to all helpful and lovely people who helped me directly or indirectly to complete this work.
Finally, my deepest thanks go to my family for their patience, constant encouragement, continuous care and support during my study.
Dedication I dedicate this work to: My lovely family and to my children (Lawk , Lalo V Khwncha)
My dear parents My loving sister and brother My faithful friends, the companions of the long road …
Nyaz
Abstract High quality video standards need a security level to stream their data in a communication channel among different devices (Computer, Satellite TV, iPad, Notepad and Mobile phones). Watermarking is the process of inserting secret information (watermark) into digital multimedia (image, audio and video). In watermarking robustness is the resistance of the watermark against attack such as compression, filtering or cropping. In this research work, a robust digital video watermarking based on High Efficiency Video Coding is proposed. It is aimed to design and implement a video watermarking scheme in a spatial domain. An important level of security is added through a robust watermarking scheme, which is used for hiding ownership profile inside the video. This security protects the copyright protection of the digital video contents. Finally, the proposed system can back up the video files in one of the common cloud computing systems like Zipcloud, Carbonite, Dropbox, and Mozy. The input video file is first extract into frames and then converted from RGB color space to the YUV space. The motion detection algorithm for the change detection is applied to separate the intra from inters frames. Two types of watermark are used, video profile and logo of the satellite TV, which are for embedding into intra and inter frames respectively. Finally, the watermarked video is objectively evaluated based on Peak Signal to Noise Ratio and Normalized Correlation. The HEVC coding system is used as an attack tool for video files compression. Test results show that for the different video quality (1K, 2K and 4K) when the watermark is embedded into the low pixel values the performance is getting better compared with the results obtained from high pixel values. I
The experimental results demonstrate that the watermark size also affects the performance of the proposed scheme. The PSNR value is high when the watermark size is small (30x30). The similarity (NC) between the original and the extracted watermark is high (200x200) when the watermark size is large after performing the compression attack using HEVC standard.
II
List of Abbreviations Abbreviation
Meaning
AVC
Advanced Video Coding
BCR
Bit Correct Rate
BER
Bit Error Rate
CTU
Code Tree Unit
CU
Coding Unit
ED
Euclidean Distance
GRL
Gradient based R-Lambda
HEVC
High Efficiency Video Coding
HD
High Definition
HL
High Low
HM
HEVC Test Model
HVS
Human Visual System
IEC
International Electro-technical Commission
ISO
International Standard Organization
ISRC
International Standard Recording Code
ITU
International Telecommunication Union
JCT-VC
Joint Collaborative Team on Video Coding
LBP
Local Binary Pattern
LH
Low High
LSB
Least Significant Bit
MOS
Mean Opinion Score
MPEG
Moving Picture Experts Group III
MSE
Mean Squared Error
NC
Normalized Correction
PN
Pseudo random Noise
PSNR
Peak Signal to Noise Ratio
PU
Prediction Unit
QP
Quantization Parameters
QTCs
Quantized Transform Coefficients
R-D
Rate-Distortion
RGB
Red-Green-Blue
SSE
Sum of Square Error
SSIM
Structural Similarity Index
TU
Transform Unit
VCEG
Video Coding Expert Group
WVGA
Wide Video Graphics Array
IV
Contents Abstract
I
List of Abbreviations
III
Contents
V
List of Figures
IX
List of Tables
XIII
Chapter One: General Introduction 1.1 Introduction……………………………………………………………..
2
1.2 System Components……………………………………………………. 4 1.2.1 Input Video Files..………………………………………………….. 4 1.2.2 Embedding Watermark …...………………………………………... 4 1.2.3 Attacking tool (HEVC)……...……………………………………...
4
1.2.4 Watermark Extraction ……………………………………………… 5 1.2.5 Robustness Performance Evaluation ………………………………. 5 1.2.6 Backup Watermarked Video Files………………………………….
5
1.3 Literature Survey……………………………………………….………. 5 1.4 The Problem Statement………………………………………….……...
9
1.5 The Aims of the Thesis……………………………………………..…... 9 1.6 Thesis Layout………………………………………………….………..
10
Chapter Two: Background of Watermarking based Video Codding 2.1 Introduction……………………………………………………………..
12
2.2 HEVC……….………………………………………………………….. 14 2.2.1 Coding Process………………………..……………………………. 15 2.2.2 Decoding Process……………………..……………………………. V
16
2.3 Video Watermarking.…………………………………………………... 17 2.3.1 Watermark Generation...………...………………………………….
21
2.3.1.1 Embedding Process……….…………………………………….
21
2.3.1.2 Extraction Process...…………………………………………….
22
2.3.2 Video Watermark in Spatial Domain......…………………………... 23 2.3.2.1 Least Significant Bit Modification……………………………... 24 2.3.2.2 Correlation-Based Techniques…..……………………………...
25
2.3.3 Semi-Blind Technique…………..………………………………….. 25 2.3.4 Video Watermarking Performance Parameters…………………….. 26 2.3.4.1 Imperceptibility…………………………………………………
26
2.3.4.2 Mean Squared Error(MSE)……………………………………..
27
2.3.4.3 Euclidean Distance(ED)………………………………………... 27 2.3.4.4 Peak-Signal-to-Noise-Ratio…………………………………….. 27 2.3.4.5 Normalized Correction…………………………………………. 28 2.3.4.6 Robustness……………………………………………………… 28 2.4 HEVC as Attacker Tool……………...…………………………………. 29 2.5 Backup in Cloud System……………………………………………….. 29
Chapter Three: Proposed System Design and Implementation 3.1 Introduction……………………………………………………………..
32
3.2 General Framework Layout………………………………………..…… 34 3.2.1 Change Detection (Inter and Intra frames separations)……..……… 34 3.2.2 Color Space Conversion………………...………………………….. 36 3.3 Watermark Embedding Process...………………………………………
38
3.4 Watermark Extraction Process..………………………………………... 41 3.5 HEVC as Attacker Tool..……………………………………………….. 44
VI
Chapter Four: Test Results Evaluation 4.1 Introduction……………………………………………………………..
47
4.2 Test Sequence…………………………………………………………... 48 4.2.1 Video Sample - 1K…………………………………………………. 48 4.2.2 Video Sample - 2K…………………………………………………. 49 4.2.3 Video Sample - 4K…………………………………………………. 50 4.3 Involved Parameters...………………………………………………….. 51 4.3.1 Change Detection………….………………………………………..
51
4.3.2 Watermark Size and Scale Factor…...……………………………… 52 4.4 Watermark Embedding Process………….……………………………... 53 4.4.1 Video Sample-1K…………….…………………………………….. 53 4.4.2 Video Sample-2K..…………………………………………………. 54 4.4.3 Video Sample-4K……………………………….………………….. 55 4.5 Watermark Extraction Process………………………………………….
56
4.5.1 Video Sample-1K…………….…………………………………….. 56 4.5.2 Video Sample-2K..…………………………………………………. 61 4.5.3 Video Sample-4K……………………………….………………….. 66 4.5.4 Embedding and Extracting Watermark from Intra Frame………….. 71 4.6 Watermarked Video Attack…………………………………………….. 72 4.5.1 Video Sample – 1K…………………………………………………
72
4.5.2 Video Sample – 2K…………………………………………………
81
4.5.3 Video Sample – 4K…………………………………………………
90
4.7 Performance Evaluation………………………………………………...
99
4.6.1 Video Sample – 1K…………………………………………………
99
4.6.2 Video Sample – 2K…………………………………………………
101
4.6.3 Video Sample – 4K…………………………………………………
103
VII
Chapter Five: Conclusions and Future Works 5.1 Conclusions……………………………………………………………..
112
5.2 Suggestions for Future Work…………………………………………...
113
References………………………………………………………………….. 115
VIII
List of Figures Figure No.
Figure Name
Page No.
1.1
General diagram of the video watermarking scheme……………
3
2.1
General structure of an HEVC encoder and decoder…................
14
2.2
HEVC coding process…..……………………………………….
15
2.3
HEVC decoding process………………………………………...
16
2.4
General digital watermarking life cycle…………………………
17
2.5
Cryptography techniques...………………………………………
18
2.6
Steganography system steps…..…………………………………
19
2.7
Digital image watermarking types..……………………………...
21
2.8
Watermark embedding unit……………………………….……..
22
2.9
The extraction/detection process……………...…………………
23
2.10
Semi-blind watermark embedding scheme..…………………….
26
3.1
General structure of the proposed system……………………….
33
3.2
Flowchart of Intra frame depiction…………………………..…..
34
3.3
Flowchart of Inter frames depiction..............................................
35
3.4
Change detection for 1000 frames……..……………….………..
35
3.5
The composition of digital image (RGB and YUV)…….……....
36
3.6
RGB color conversion to YUV…………….……………………
37
3.7
Color conversion pseudo code……………………………….......
37
3.8
Watermark embedding process flowchart for intra frame……….
39
3.9
Watermark embedding process flowchart for inter frames.……..
40
3.10
Watermark extraction process flowchart for intra frame………...
42
3.11
Watermark extraction process flowchart for inter frames……….
43
3.12
Compression attack using HEVC………………………………..
44
IX
4.1
HoneyBees (1280x720)………...………………..........................
48
4.2
BigBuckBunny (1920x1080)……………………………..….......
49
4.3
HoneyBees (4096x2304)……...………..………..........................
50
4.4
Similarity versus frame number…………….…………………...
51
4.5
Different watermark size………………………………………...
52
4.6
1K – Embedding watermark in high pixel values...……………..
53
4.7
1K – Embedding watermark in low pixel values....……………..
53
4.8
2K – Embedding watermark in high pixel values...……………..
54
4.9
2K – Embedding watermark in low pixel values....……………..
54
4.10
4K – Embedding watermark in high pixel values...……………..
55
4.11
4K – Embedding watermark in low pixel values....……………..
55
4.12
Extract watermark – 1K………………………………………….
56
4.13
Extract watermark – 1K………………………………………….
57
4.14
Extract watermark – 1K………………………………………….
57
4.15
Extract watermark – 1K………………………………………….
58
4.16
Extract watermark – 1K………………………………………….
59
4.17
Extract watermark – 1K………………………………………….
59
4.18
Extract watermark – 1K………………………………………….
60
4.19
Extract watermark – 1K………………………………………….
61
4.20
Extract watermark – 2K………………………………………….
61
4.21
Extract watermark – 2K………………………………………….
62
4.22
Extract watermark – 2K………………………………………….
63
4.23
Extract watermark – 2K………………………………………….
63
4.24
Extract watermark – 2K………………………………………….
64
4.25
Extract watermark – 2K………………………………………….
65
4.26
Extract watermark – 2K………………………………………….
65
4.27
Extract watermark – 2K………………………………………….
66
X
4.28
Extract watermark – 4K………………………………………….
67
4.29
Extract watermark – 4K………………………………………….
67
4.30
Extract watermark – 4K………………………………………….
68
4.31
Extract watermark – 4K………………………………………….
68
4.32
Extract watermark – 4K………………………………………….
69
4.33
Extract watermark – 4K………………………………………….
69
4.34
Extract watermark – 4K………………………………………….
70
4.35
Extract watermark – 4K………………………………………….
71
4.36
Text to binary conversion………………………………………..
71
4.37
Extracted watermark after attack for (30x30) logo size…………
72
4.38
Extracted watermark after attack for (50x50) logo size…………
73
4.39
Extracted watermark after attack for (60x60) logo size…………
74
4.40
Extracted watermark after attack for (200x200) logo size………
76
4.41
Extracted watermark after attack for (30x30) logo size…………
77
4.42
Extracted watermark after attack for (50x50) logo size…………
78
4.43
Extracted watermark after attack for (60x60) logo size…………
79
4.44
Extracted watermark after attack for (200x200) logo size………
80
4.45
Extracted watermark after attack for (30x30) logo size…………
81
4.46
Extracted watermark after attack for (50x50) logo size…………
82
4.47
Extracted watermark after attack for (60x60) logo size…………
83
4.48
Extracted watermark after attack for (200x200) logo size………
85
4.49
Extracted watermark after attack for (30x30) logo size…………
86
4.50
Extracted watermark after attack for (50x50) logo size…………
87
4.51
Extracted watermark after attack for (60x60) logo size…………
88
4.52
Extracted watermark after attack for (200x200) logo size………
89
4.53
Extracted watermark after attack for (30x30) logo size…………
90
4.54
Extracted watermark after attack for (50x50) logo size…………
91
XI
4.55
Extracted watermark after attack for (60x60) logo size…………
92
4.56
Extracted watermark after attack for (200x200) logo size………
94
4.57
Extracted watermark after attack for (30x30) logo size…………
95
4.58
Extracted watermark after attack for (50x50) logo size…………
96
4.59
Extracted watermark after attack for (60x60) logo size…………
97
4.60
Extracted watermark after attack for (200x200) logo size………
98
4.61
Effect of logo size and QP on PSNR…………………………….
99
4.62
Effect of logo size and QP on PSNR…………………………….
100
4.63
Effect of logo size and QP on PSNR…………………………….
101
4.64
Effect of logo size and QP on PSNR…………………………….
102
4.65
Effect of logo size and QP on PSNR…………………………….
103
4.66
Effect of logo size and QP on PSNR…………………………….
104
4.67
PSNR before and after attack – high pixel value………………..
106
4.68
PSNR before and after attack – low pixel value…………………
106
4.69
NC values for 1K video………………………………………….
107
4.70
NC values for 2K video………………………………………….
107
4.71
NC values for 4K video………………………………………….
108
4.72
PSNR values for 1K video………………………………………
108
4.73
PSNR values for 2K video………………………………………
109
4.74
PSNR values for 4K video………………………………………
109
XII
List of Tables
Table No.
Table Name
Page No.
3.1
QP effect on the video size…………………………………………
45
4.1
Test sequence (HoneyBees)………………………………………...
48
4.2
Test sequence (BigBuckBunny)..…………………………………..
49
4.3
Test sequence (HoneyBees)………………………………………..
50
4.4
Effect of threshold for separate Intra from Inter frames……………
51
4.5
Effect of watermark size…………..………………………………..
52
4.6
NC versus scale factor……………………………………………...
56
4.7
NC versus scale factor……………………………………………...
57
4.8
NC versus scale factor……………………………………………...
57
4.9
NC versus scale factor……………………………………………...
58
4.10
NC versus scale factor……………………………………………...
58
4.11
NC versus scale factor……………………………………………...
59
4.12
NC versus scale factor……………………………………………...
60
4.13
NC versus scale factor……………………………………………...
60
4.14
NC versus scale factor……………………………………………...
61
4.15
NC versus scale factor……………………………………………...
62
4.16
NC versus scale factor……………………………………………...
62
4.17
NC versus scale factor……………………………………………...
63
4.18
NC versus scale factor……………………………………………...
64
4.19
NC versus scale factor……………………………………………...
64
4.20
NC versus scale factor……………………………………………...
65
4.21
NC versus scale factor……………………………………………...
66
4.22
NC versus scale factor……………………………………………...
66
XIII
4.23
NC versus scale factor……………………………………………...
67
4.24
NC versus scale factor……………………………………………...
68
4.25
NC versus scale factor……………………………………………...
68
4.26
NC versus scale factor……………………………………………...
69
4.27
NC versus scale factor……………………………………………...
69
4.28
NC versus scale factor……………………………………………...
70
4.29
NC versus scale factor……………………………………………...
70
4.30
NC for different QP – logo size 30x30…………...………………...
72
4.31
NC for different QP – logo size 50x50…………...………………...
73
4.32
NC for different QP – logo size 60x60…………...………………...
74
4.33
NC for different QP – logo size 200x200………….......…………...
75
4.34
NC for different QP – logo size 30x30…………...………………...
76
4.35
NC for different QP – logo size 50x50…………...………………...
77
4.36
NC for different QP – logo size 60x60…………...………………...
78
4.37
NC for different QP – logo size 200x200………….......…………...
79
4.38
NC for different QP – logo size 30x30…………...………………...
81
4.39
NC for different QP – logo size 50x50…………...………………...
82
4.40
NC for different QP – logo size 60x60…………...………………...
83
4.41
NC for different QP – logo size 200x200………….......…………...
84
4.42
NC for different QP – logo size 30x30…………...………………...
85
4.43
NC for different QP – logo size 50x50…………...………………...
86
4.44
NC for different QP – logo size 60x60…………...………………...
87
4.45
NC for different QP – logo size 200x200………….......…………...
88
4.46
NC for different QP – logo size 30x30…………...………………...
90
4.47
NC for different QP – logo size 50x50…………...………………...
91
4.48
NC for different QP – logo size 60x60…………...………………...
92
4.49
NC for different QP – logo size 200x200………...………………...
93
XIV
4.50
NC for different QP – logo size 30x30…………...………………...
94
4.51
NC for different QP – logo size 50x50…………...………………...
95
4.52
NC for different QP – logo size 60x60…………...………………...
96
4.53
NC for different QP – logo size 200x200………...………………...
97
4.54
PSNR for different QP and Logo Size……………………………...
99
4.55
PSNR for different QP and Logo Size……………………………...
100
4.56
PSNR for different QP and Logo Size……………………………...
101
4.57
PSNR for different QP and Logo Size……………………………...
102
4.58
PSNR for different QP and Logo Size……………………………...
103
4.59
PSNR for different QP and Logo Size……………………………...
104
4.60
NC versus video sample – high pixel value………………………...
105
4.61
NC versus video sample – low pixel value.………………………...
105
XV
Chapter 1
General Introduction
Chapter One General Introduction 1.1 Introduction The High Efficiency Video Coding (HEVC) is a new standard for video compression that provides a better performance in terms of compression ratio and quality of video reconstruction compared to the previous standards. Standards that exist for video coding were based on both spatial and transform domains including (H.261 which is represented by MPEG1 standard, H.262/MPEG2, H.264/MPEG4 and Advanced Video Coding (AVC)). HEVC is applied on different applications such as High Definition TV broadcasting, Network streaming security, real time voice and video conferencing and telemedicine system [Sal, 14]. Digital video watermarking plays a crucial role for proofing digital copyright protection for the above mentioned video standards. The information hiding main parameters is payload (capacity), security and robustness. The need of the secure communication channel and video data transfer has rapidly increased with progressing for emerged new technologies. The main technique that is used for multimedia data protection (copyright) is digital watermarking. It can be used for embedding different media types into text, image, audio and video as well. The process of the embedding watermark into the digital video may cause a remarkable distortion into the visible components of the host watermarked signal. If the watermark cannot simply extract from the watermarked signal even after subjecting to the some watermarking attacks (geometrics, filters and compression) then it is considered as robust embedding [Rin, 11]. 2
Chapter 1
General Introduction
The robust watermarking scheme consists of three phases starting with the embedding process then attacking the watermarked video and finally extracting it. The performance of the proposed scheme is evaluated using objective quality measure Peak Signal to Noise Ratio (PSNR) between original and watermarked video and a similarity measurement (NC) between original and extracted watermark. Figure (1.1) represents the general diagram for any robust video watermarking system.
Input Video
Extracting frames
Find Similarity
Intra frames
Inter frames
Color Conversion into YUV
Color Conversion into YUV
Add Logo
Add Text
Watermarked Intra frame
Watermarked Inter frames
Apply HEVC Attack
Extract Text and Logo
Evaluate the robustness using PSNR and Normalized Correlation Fig (1.1) General diagram of the video watermarking scheme 3
Chapter 1
General Introduction
1.2 System Components The proposed robust video watermarking algorithm based on HEVC coding standard for backup media files has the following basic components.
1.2.1 Input Video Files The video files that are used for the test are based on the real time capturing from the webcam or any external cameras and the off line video files coded by HEVC standard. The color system is in RGB format which is the standard for color video files ranging from the low resolution like (MPEG1,2) to the HD resolution such as (1K, 2K, 4K) used by H.265 coding system.
1.2.2 Embedding Watermark Two types of watermark are used for embedding both Intra and Inter frames based on the correlation between frames according to the given threshold. After separating Intra frames, the first watermark which represents the profile of the company or a TV channels (Text watermark) is added to the first detected Intra frame. The second watermark that is an image of the company logo is embedded to the detected Inter frames.
1.2.3 Attacking tool (HEVC) The watermarked video files are compressed using HEVC as an attacker tool to measure the robustness of the proposed scheme. Different quantization parameters (QP) are tested in the x.265 encoder software to get better compression and high robustness with preserving the quality of the video.
4
Chapter 1
General Introduction
1.2.4 Watermark Extraction The extraction step is the process of retrieving the watermark bits in both intra and inter frames. It then recovers the text and logo watermark from the watermarked video.
1.2.5 Robustness Performance Evaluation To evaluate the robustness of the watermarking method, the extracted watermark is compared to the original one by using similarity metric. The original video file quality is also matched with the watermarked video using PSNR metric. The video file sizes are also taken into consideration for the backup purpose.
1.2.6 Backup Watermarked Video Files One of the common existing cloud computing systems such as (Drop Box, One Drive, Google Drive …etc.), is chosen to backup the compressed watermarked video.
1.3 Literature Survey Several studies and research about the digital video watermarking and video compression, and have been published within the last ten years. Some of them are listed below: 1. Jamal Hussein and Aree Mohammed [Jam, 09] presented a robust video watermarking scheme implemented in both spatial and wavelet domains. The embedding process is done by exploiting the motion estimation approach in the video. The method utilizes the HL and LH bands to embed the watermark where the motion does not affect the quality of the extracted watermark if the video is subjected to different types of malicious attacks. 5
Chapter 1
General Introduction
The test results show that the similarity between the original and the attacked watermark is almost the same in frequency domain. 2. Manish K Thakur et al [Man, 10] proposed a performance evaluation through objective quality measurement with available digital video watermarking techniques. Two different metrics, PSNR and Structural Similarity Index SSIM, are used to show the performance. Experimental results show that the SSIM are more accurate as compared with the results obtained during subjective quality analysis. 3. Radu Ovidiu Preda and Nicolae Vizireanu [Rad, 11] developed a novel digital video watermarking method for copyright protection in a spatial domain. The binary images are used as a watermark and it is embedding in the block of luminance band (Y channel) by quantization strategy. The original image and watermark are not used in the extraction process and it is considered as a blind technique. Proposed algorithm is tested the robustness of the algorithm against seven different attacks. Experimental results show that the embedded watermark is invisible and robust to the attacks. 4. Jani Lainema et al [Jan, 12] described features contributing to raise the compression efficiency involving a quadtree-based variable block size coding structure. The design principle is applied during the development of the new intra coding methods and analyzed the compression performance of the individual tools. They found out the significant subjective picture quality improvement which is also reported when comparing the result picture at fixed bitrate. 5. Gary J. Sullivan et al [Gar, 12] introduced the HEVC standardization effort to enable significantly the improved compression performance relative to the existing standards. It is also used for the reduction of the range of the 6
Chapter 1
General Introduction
bitrate equally in perceptual video quality. In this paper, an overview for the technical features and characteristics of the HEVC algorithm is provided. 6. Jens-Rainer Ohm et al [Jen, 12] presented and applied a unified approach to the analysis of design including H.262/MPEG-2, H.263, H.264/MPEG4(AVC), and HEVC. The result of the test for WVGA and HD sequence shows that HEVC encoders can achieve equivalent subjective reproduction quality that follows to H.264/MPEG-4 AVC when using %50 less bitrate. The HEVC design is effective particularly for low bit rate, high resolution video content and low delay for communication applications. 7. Yimin Zhou [Yim, 13] proposed a rate control algorithm which is a combination of a linear rate-distortion (R-D) model, a PID buffer feedback controller, and an incremental computation method to quantization parameter (QP). The proposed scheme is compared with fixed QP encoding on the HM5.1 with the same configuration. An accuracy of buffer control is obtained by avoiding the buffer overflow and underflow, and obtained a smooth visual quality. The proposed method is the first algorithm that applies to the Intra rate controller in HEVC standard. 8. Salahuddin Swati et al [Sal, 14] developed an algorithm that has a high potential of using in applications such as broadcast and hiding of metadata. The process of embedding the watermark data is located into the Quantized Transform Coefficient (QTCs) during the encoding process. The watermark is completely extracted in the decoding process. Test results show that the modified scheme dose not significantly change the video quality, nor does it increase the bitrate. 9. Aisheng Yang et al [Ais, 14] described a novel video coding standard (HEVC) that optimizes its coding efficiency in terms of the sum of square 7
Chapter 1
General Introduction
error (SSE). HEVC does not take into account the perceptual properties of the test video files and thus it is not efficient for perceptual video codec. To handle this issue, an effective perceptual rate control algorithm for HEVC is proposed based on the human visual system (HVS) inspection that the region with less sensitive perception can to allow more distortion. Simulation test results show that the proposed scheme is able to outstandingly improve the performance of the rate distortion perception, compared with the original version of HEVC. 10. Miaohui Wang et al [Mia, 15] proposed the HEVC encoder that assumes a new R-lambda based on rate control model. The aim of the method is to reduce the estimated bit error. In other hand, the R-lambda model fails to take into account the frame-content complexity that essentially deteriorates the performance evaluation of the bit rate control. In this paper, a gradient based R-lambda (GRL) model is developed for the intra frame rate control, where the gradient can considerably measure the frame-content complexity and improve the performance of the traditional R-lambda method. Test results show that the proposed GRL method can effectively minimize the estimated bit error and lead the video quality in HEVC better for all intra frame coding. 11. Jong-Hyeok Lee et al [Jon, 15] introduced a fast video coding algorithm to reduce the run time computation of the HEVC RExt encoder. To design and implement the intra mode decision algorithm, a texture of based feature of the current prediction unit is extracted using Local Binary Pattern (LBP). Test results show that the time computation of the encoding process can be minimized by up to 12.35% on average in the AI-Main profile configuration with only a small bit-rate increasing and a PSNR decreasing,
8
Chapter 1
General Introduction
in comparison with the HEVC test model (HM) 12.0-RExt4.0 original software. 12. Ali A.Elrowayati et al [Ali, 16] implemented a new approach for HEVC that gives a high robustness when it is subjected to the noise channel errors. The proposed method increases the detection rate error in the successive frames inside video sequences which are transmitted over noisy communication channels. The modified scheme dose not significantly change the video quality, nor does it increase the bitrate. Test results show that the proposed method yields better robustness against noise channel while preserving good quality for extracted watermark. Another contribution of this work is to improve the performance through the error detection and correction. The watermark can recover almost completely when the frames dropping rate in watermarked video is less than 20%.
1.4 The Problem Statement The motivation for choosing the topic of “video watermarking” in this research work is to secure the communication channel for transmitting video data with a high quality (1K,2K,4K). As the HEVC is a new emerging coding standard for video compression with a high quality (HD), it must be backing up as much secure as possible. In this context a robust video watermarking scheme is proposed to solve the above problems.
1.5 The Aims of the Thesis The aim of this research work is to develop and implement an efficient video file backup system in cloud storage with a video compression based on HEVC which is an emerging video compression standard that provides better compression performance as compared to its predecessor (i.e. H.264/AVC). 9
Chapter 1
General Introduction
A level of security is added to the proposed scheme by embedding the ownership profile and the logo of the specific TV channel. The mentioned security protects the copyright protection of the digital video contents. The common cloud computing system examples are Zipcloud, Carbonite, Dropbox and Mozy which are used for the proposed system in order to backup a video files.
1.6 Thesis Layout Apart from the current chapter, the rest of the thesis is organized in four chapters, which are: Chapter Two is entitled “Background of Watermarking based Video Codding”. This chapter presents the basic components of the HEVC standard and describes some digital video watermarking schemes. Finally, HEVC is used as attacker tools. Chapter
Three
is
entitled
“Proposed
Scheme
Design
and
Implementation”. In this chapter the detail of the proposed methods is shown through the algorithm, flowchart and block data diagram. Chapter Four is entitled “Test Results Evaluation”. This chapter presents the conducted test results to evaluate the proposed scheme using the performance parameters (PSNR and Auto Correlation). The robustness is also given after attacking the video files with HEVC tool. Chapter Five is entitled “Conclusions and Future Works”. It is devoted to list out some of the conclusions derived from the test results. Moreover, some suggestions are proposed for future directions.
10
Chapter 2
Background of Watermarking based Video Coding
Chapter Two Background of Watermarking based Video Codding 2.1 Introduction Digital watermarking is a kind of marker or a pattern of bits secretly embedded information into a noise-tolerant digital signal such as an audio, video or image data. It is usually used to recognize ownership of the copyright information of such signal. The digital signal may carry more than one nonidentical watermark at a time. It is a method that is related to steganography, such that they hide a message inside a digital signal [Alk, 14]. In every watermarking system, there are many of important useful properties. Some of these preferable properties are often conflicted and they are often forced to accept some tradeoffs between these properties determined by the application of the digital watermarking process. The most important property is the effectiveness which is the probability that the message in a digital watermarked media will be correctly detected [Kar, 14]. The other important property is the imperceptibility of the watermark and the last property is the payload size. The first step of the watermarking process is the embedding of a message inside the cover signal (video in this work). The size of this message is commonly important as the systems need a relatively huge payload to be embedded [Gar, 12]. HEVC is comparatively a novel video compression standard. The main objective of HEVC standard is to enhance video resolutions and to manipulate parallel processing construction. It is used for some kind of applications such as broadcast of high definition (HD) TV signals, video content acquisition and 12
Chapter 2
Background of Watermarking based Video Coding
editing systems, security applications, camcorders, Blue-ray discs, internet and mobile network video and real-time conversational applications that include video conferencing, video chat, and tele-presence systems [Sal, 14] HEVC standard is the more dislike video project of the ITU-T Video Coding Experts Group (VCEG) and the ISO/IEC Moving Picture Expert Group (MPEG)
standardization
firms,
in operation with cooperation
Joint
Collaborative Team on Video Coding (JCT-VC). The initial edition of the standard is anticipated to be established in January 2013[Hsu, 15]. The objective of the most video coding standards is to optimize coding efficiency. Coding Efficiency is to reduce the bit rate needed for video content and extend the level of quality of the video. On the other hand, it is formulated to increase the quality of the obtained video within a particular available bit rate [Jen, 12]. Cloud computing is the resource that may cover application such as programming interface, servers, storages, applications, blogs, presentations, emails, documents, chats, software, and networks. It gives the secure access for the users to such resources by taking benefit of self-service and whenever cloud computing technologies by their smart phones, pad tablets, notebooks, laptops, and computers. There are many cloud computing backup providers which have vendor provided various backup scheme for users [Raj, 13]. In this chapter, the general overview about the digital video watermarking and the emerging coding standard (HEVC) are given. In the next section, a detail description of the coding and decoding process of the HEVC is explained. Then the watermark embedding and extraction process are given in section (2.3). The implementation of the proposed system in spatial domain is indicated in section (2.3.2). Finally the attacker tool which is the standard HEVC and the backup system are given in section (2.4 and 2.5) respectively.
13
Chapter 2
Background of Watermarking based Video Coding
2.2 HEVC HEVC is a hybrid video compression algorithm that constructs on the Intra/Inter frame prediction and 2D DCT transform. Similar to H.264, HEVC acquires a block based hybrid coding structure. The development of the coding efficiency is executed by utilizing many recent techniques, like Coding Unite (CU) with variable block-size varying from 4x4 to 64x64, Prediction Unit (PU), Transform Unit (TU), quad tree constrictions and advance motion vector estimation. The HEVC quiet optimizes the coding performance in terms of the Sum of Squared Error (SSE). The video quality is eventually decided by eyes; it is very crucial to involve the perceptual vision of HEVC with the aim of increasing the perceptual quality under a given bit-rate [Gar, 12]. The goal of the rete control is to get the optimal video quality below constraints of the bandwidth channel and codec-buffer in an accurate way. This is initiate the rate distortion/quantization algorithm based on the characteristics of remaining or input the video and then uses this algorithm to decide the acceptable Quantization Parameter (QP) [Ais, 14]. Figure (2.1) shows the general block diagram of HEVC coding and decoding process.
Fig. (2.1) General Structure of an HEVC encoder and decoder [VCO, 15] 14
Chapter 2
Background of Watermarking based Video Coding
2.2.1 Coding Process HEVC was designed initially to improve the coding efficiency compared to H.264/MPEG-4 and AVC. It reduces the bitrate requirements by half of the original size of the frame with preserving the acceptable level of image quality, at the expense of increased computational complexity [Bog, 15]. HEVC was implemented with the goal of allowing video content to have a double data compression. Depending on the application requirements, HEVC encoders can make a tradeoff of the computational complexity, compression rate, robustness to errors, and encoding delay time [Viv, 14]. In figure (2.2) the coding process part of HEVC is presented.
Fig. (2.2) HEVC coding process [sal, 14]
15
Chapter 2
Background of Watermarking based Video Coding
The performance of the coding efficiency of any video coding standard is evaluated by using objective and subjective metrics, such as PSNR, or Mean Opinion Score (MOS). Subjective measurement of video quality is considered as an important method to measure the video quality since humans perceive video quality subjectively. The HEVC video coding layer uses the same hybrid method used in all modern video standards which starts from H.261, it uses inter/intra picture prediction and 2D DCT transform. An HEVC encoder proceeds by splitting a picture into block shaped regions for the first picture, or the first picture of a random access which uses intra picture prediction. Intra-picture prediction which is the prediction of blocks in the picture is based only on the information in that picture. For other pictures, inter picture prediction is used, while the prediction information is used from other pictures. After the prediction scheme is finished and the picture goes through the loop filters, the final picture representation is stored in the decoded picture buffer [Jan, 12]. 2.2.2 Decoding Process The inverse process of the coding is depicted in figure (2.3). Intra Frame Inverse Quantizatio n Encoded Bit Stream
Inverse DCT
Inter Frame
DE blocking Filter Sample Adaptive Offset Filter
Inverse VLC
Reconstructed Frame Motion Compensatio n
Motion Vectors
Fig. (2.3) HEVC decoding process
16
Previous Frame Memory
Chapter 2
Background of Watermarking based Video Coding
2.3 Video Watermarking Video Watermarking is a new and rapidly developed field in the area of multimedia. There are some factors that are involved towards the activation of interest in multimedia field. The society is polluted by the enormous privacy of digital data, as copying of digital media become easy. It is a period which needs
to
appear
for
fight
against
“Intellectual
property
rights
infringements”[Rin, 11]. Digital copyright protection must not be destructed caused by harmful attacks. Alteration of the digital data needs to be hidden at some point [See, 14]. Figure (2.4) shows the life cycle of digital watermarking.
Fig. (2.4) General digital watermarking life cycle
There are various approaches for information hiding into digital media. They are used for multi-purposes as well as copyright protection. There are two basic procedures of information hiding such as steganography, watermark and fingerprint. The idea of digital watermarking has essentially been taken from the steganography. The steganography means cover writing and cryptography means secret writing. The message is encrypted before transmission and decrypted at the recipient end with the help of a key. No one can access the content without having the original key. The encrypted message is called the cipher text and the message is called the plain text. The information is protected before transmission. However, after decryption, the information becomes unprotected and it can be copied and distributed [Rin, 11]. 17
Chapter 2
Background of Watermarking based Video Coding
Visibility is a term associated with the perception of the human eye. A watermarked
image
in
which the
watermark
is
imperceptible,
or
the watermarked image is visually identical to its original constitutes an invisible digital watermarking. Examples include images distributed over internet with watermarks embedded in them for copyright protection. Those, which fail, can be classified as visible digital watermarks. Examples include logos used in papers in currencies. The encryption and decryption mechanism of a given message is described in figure (2.5).
Fig. (2.5) Cryptography Techniques
In steganography method, the message is embedded into the digital media rather than encrypting it in such a way that nobody expects the sender and the intended recipient can even realize that there is a hidden message. The digital 18
Chapter 2
Background of Watermarking based Video Coding
media content called the cover can be determined by anyone but the message hidden in the cover can be detected by only the person having the actual key. After the message is embedded into the cover image, the cover image becomes stego image. Thus steganography actually relates to covering point-to-point communication between two parties. That is steganography methods are usually not robust against modification of the data or have only limited robustness [Ger, 00]. Steganography is the art and science of hiding messages. Steganography and cryptology are similar in the way that they both are used to protect important information. The difference between the two is that Steganography involves hiding information so it appears that no information is hidden at all. Figure (2.6) shows the steganography steps.
Fig. 2.6: Steganography system steps
According to the human perception, the watermarking techniques can be divided into three parts:
A visible watermark is a semi-transparent type added into an image and is
visible to the viewer. Visible watermarking is used to determine the ownership and copyright protection. It is absolutely the information text or a logo that clearly recognizes the owner of the image and it’s copyright. The 19
Chapter 2
Background of Watermarking based Video Coding
size of the message is less important than its content. Visible watermarks are more robust against image transformation (especially if you use a semitransparent watermarking placed over a whole image). Thus, they prefer robust copyright protection of intellectual property than in digital format.
An invisible watermark is embedded into the digital data in such a way
that the changes made to the pixel values are perceptually not noticed. Invisible watermark is used as evidence of ownership and to detect appropriate digital data. Invisible watermark is an embedded image, which cannot be recognized by human’s eyes. Only electronic devices (or specialized software) can extract the hidden information to identify the copyright owner. They are also used to mark a specialized digital content (text, images or even audio content) to prove its authenticity. The copyright protection is the main area of using digital watermarks; they are also used for such application as advertising (adding logo company name as a watermark for promotion rather than for protection) or even adding memo to digital photos title. Visibility is a term associated with the perception of the human eye. A watermarked image
in which the watermark is imperceptible, or
the watermarked image is visually identical to its original constitutes an invisible digital watermarking. Examples include images distributed over internet with watermarks embedded in them for copyright protection. Those, which fail, can be classified as visible digital watermarks. Examples include logos used in papers in currencies.
Dual watermark is the combination of both visible and invisible
watermark. An invisible watermark is used as a backup for the visible watermark [Vip, 11]. Figure (2.7) illustrates the digital watermarking types.
20
Chapter 2
Background of Watermarking based Video Coding
a. Visible watermark
b. Invisible watermark
Fig. (2.7) Digital image watermarking types
2.3.1 Watermark Generation Watermark can be generated as a bit stream (0, 1) for different types of signals such as text message, audio, image and video. For image copyright applications, the common practice is that it should be around 60-80 bits. This is necessary to embed the International Standard Recording Code (ISRC) years of copyright and the granted permissions on the work [Jam, 09].
2.3.1.1 Embedding Process
Let us denote an original image by I, a watermark by W. The watermarked image by
and K is the embedded key (see Figure 2.8). The embedding
function
takes on its input the image I, watermark W and key K and
generates a new watermarked image, denoted with
. Introduction of the
embedded key K is necessary for enhancing the security aspect of the watermarking system. Before the embedding process, the original image can be either transformed in the frequency domain or the embedding can be performed in spatial domain. The domain selection depends on the selected watermarking technique. If the embedding is performed in frequency domain, the inverse transform must be applied in order to obtain the watermarked
21
Chapter 2
Background of Watermarking based Video Coding
image [Nit, 13]. Mathematically expressed, the embedding function for the spatial domain techniques can be represented as follows: (
….. (2.1)
)
For the frequency domain technique, the following expression is valid. (
….. (2.2)
)
Figure (2.8) depicted the watermark embedding unit as a part of watermarking system.
Fig. (2.8) Watermark embedding unit
2.3.1.2 Extraction Process A detector function determined. The image
takes an image
whose ownership is to be
can be a watermarked or an un-watermarked image.
In a general case, it can also be an altered image. The detector function either recovers a watermark
from the image or checks the presence of the
watermark W in a given watermarked image
. In this procedure the same
key K is used. In this process the original image I can also be included, which depends on the selected watermarking scheme [Kum, 06]. Mathematically expressed, the extraction procedure for semi-blind extraction (extraction without using the original image I) can be expressed as follows: (
)
….. (2.3)
For non-blind extraction (extraction using the original image) the following holds: 22
Chapter 2
Background of Watermarking based Video Coding
(
)
….. (2.4)
Figure (2.9) Shows the extraction/detection process for still images is presented and used for further explanation purposes.
Fig. (2.9) the extraction/detection process
2.3.2 Video Watermark in Spatial Domain Video watermarking techniques can be grouped into two categories; spatial-domain and frequency-domain techniques. Spatial-domain approaches embed a digital watermark in the frames of a given from video by altering its pixels directly.
These
methods
are easy to implement and need few
computational resources, however, they are not robust versus common digital signal processing operations such as video compression. The transform domain watermarking approaches alter the coefficients of the transformed video frames according to a pre-determined embedding plan. The scheme broadcasts the watermark in the spatial domain of the video frame, therefore making it very hard to remove the embedded watermark. Compared to spatial domain, frequency domain techniques demonstrate to be more successful and powerful with regard to achieving the imperceptibility and robustness of digital watermarking scheme [Rad, 12]. 23
Chapter 2
Background of Watermarking based Video Coding
The watermark design and insertion method do not include any transformation. Some simple techniques like adding or replacement are used for the integration of watermark with the current signal and embedding occurring directly in the pixel domain. The watermark is applied in pixel or coordinate-domain and the main strengths of pixel-domain system are conceptually simple and have very little computational complexities. Outcomes they have proven to be most attractive for applications of video watermarking where real time performance is a main problem. However, they also show some major limitations: The need for complete spatial synchronization guide to high susceptibility, to de-synchronization attacks needed of consideration of temporal axis in vulnerability to video processing and multiple frame collusion and watermark optimization is difficult using only spatial analysis method [Gop, 13].
2.3.2.1 Least Significant Bit Modification
This Technique is used to insert a watermark into the (LSB) of pixels that are detected in the neighbor of image outline. The LSB approach was implied on the modifications of LSB that damaged the watermark. However, the LSB methods also show some crucial limitations. Since absolute spatial synchronization is required, sensitivity to de-synchronization attacks is enlarged. Collusions of multiple frames may occur as caused by the lack of consideration of the temporal axis [Pre, 14].
24
Chapter 2
Background of Watermarking based Video Coding
2.3.2.2 Correlation-Based Techniques
Another technique method for embedding watermark is to utilize the correlation properties of addition pseudo random noise format as applied to an image. A Pseudo random Noise (PN) pattern W(x, y) is added to the cover image I(x, y), according to the equation below: (
)
(
)
(
)
….. (2.5)
K denotes a gain factor, and Iw the resulting watermarked image. Increasing k increases the robustness of the watermark at the expense of the quality of the watermarked image. To retrieve the watermark, the same pseudo-random noise generator algorithm is used with the same key, and the correlation between the noise pattern and possibly watermarked image computed. If the correlation exceeds a certain threshold T, the watermark is detected, and a single bit is set. This method can easily be extended to a multiple-bit watermark by dividing the image up into blocks [Suj, 15].
2.3.3 Semi-Blind Technique According to the digital watermarking extraction process, techniques can be divided into three types (Non-blind, Semi-blind and Blind). Non-blind watermarking method needs an original image and a secret key for watermark detection whereas semi-blind method needs secret key and watermark bit sequence for extraction. Blind method requires only secret keys for extraction [Pri, 13].
25
Chapter 2
Background of Watermarking based Video Coding
Figure (2.10) illustrates the embedding process of semi-blind watermarking techniques.
Fig. (2.10) Semi-Blind watermark embedding scheme
2.3.4 Video Watermarking Performance Parameters An evaluation metrics is required to assess the performance and the watermark security of a digital watermarking algorithm. Criteria which will analyze the watermarking method constructs on its most popular applications. Following are some functions for evaluating the performance and the security of watermarking schemes. The imperceptibility and the robustness of the extracted watermark can be determined by the following metrics [Asi, 15].
2.3.4.1 Imperceptibility
Imperceptibility implies to the quality of watermarked media as observed visually and they depend on human visual system. The embedding of digital watermarking into a cover image is not directly visible to the observer. Obviously, the distortion introduced to the digital watermarked content is caused by embedding mechanisms. It is desirable that the method used for watermarking should add minimal distortions to the digital content. For image perceptibility, popular evaluation criteria are based on mean-square error 26
Chapter 2
Background of Watermarking based Video Coding
(MSE), Euclidean distance (ED), peak signal-to-noise ratio (PSNR) and normalized correction (NC) [Ami, 13].
2.3.4.2 Mean Squared Error (MSE)
It is an algorithm to check distortions between cover and watermarked image. With the calculation of MSE we can detect the change in watermarked images. ∑ Here,
∑
denotes the cover image and
(
….. (2.6)
)
denotes the watermarked data.
2.3.4.3 Euclidean Distance (ED)
It is a common distance between two points in a Euclidean space and for images, two-dimensional ED is used. The specialized image ED methods are also presented by the basic idea that can be covered by two-dimensional ED. Following formula is used to calculate the ED between two images. (
)
∑
∑
(
(
)
))
(
….. (2.7)
2.3.4.4 Peak-Signal-to-Noise Ratio
PSNR is the most important test to check the distortion between original and watermarked image because it uses mean squared error also. PSNR can be calculated by the following formula. (
)
….. (2.8)
PSNR is usually used to check the nature of watermarked image. represents the maximum intensity of the image and the noise is the error which is calculated by
. PSNR estimates the quality of image or video 27
Chapter 2
Background of Watermarking based Video Coding
reproduction according to the human vision system. When the value of PSNR is high, it indicates the high quality of the watermarked video frames.
2.3.4.5 Normalized Correction
Normalized correction is a measure of similarity between two images as a function of a time-lag applied to one of them. Following formula is used to calculate NC for two images. ∑ √∑
∑
∑
(
( (
)
(
√∑
)
))
….. (2.9)
∑
(
)
Where, O is the original watermark and E is the extracted watermark. Additionally, Hamming distance, Bit Correct Rate (BCR) and bit error rate (BER) are used for binary images. These measures can be used to measure the accuracy of recovered watermark quantitatively. The formula of the
is as
follow: (
)
∑
(
)
(
)
∑ | (
(
)
(
)
….. (2.10) ….. (2.11)
(
(
)|
)
)
….. (2.12)
Here, Y = original image & Y’ = processed image, M = width of the image & N = height of the image Y (i, j) = pixel position at (i, j) location of Y & Y’ (i, j) = pixel position at (i, j) location of Y’. 2.3.4.6 Robustness
It is a characteristic to examine the resistance against any external attacks. In many applications, the strength of the digital watermarked image to carry out a noise is important. Check the robustness of the watermarked image or frames through the attacks by using one of the metrics discussed in section 28
Chapter 2
Background of Watermarking based Video Coding
(2.3.4). It can also measure the strength of the robustness. With the help of experiments the higher robustness means that the watermarked video has to preserve their visual quality.
Watermark robustness specifies the
watermarking structure to detect and extract the specific embedded watermark according to the techniques that used in both spatial and frequency domain. It may be applied toward data, with or without intent for you to prevent detection from the embedded watermark [Ami, 13].
2.4 HEVC as Attacker Tool The video watermarking of most common attacks are frame dropping, frame averaging, statistical analysis, and loss of compression, cropping and different signal processing and geometrical attacks. Malicious attacks: In general watermarking attack involves only single frame attacks like filtering attacks, contrast and color improvement and noise adding attack. Or statistical attacks like averaging and collision attacks. Unintentional attacks may be required to degenerations that can occur during the loss of copying, or due to compression of the video during re-encoding or because of change of frame rate and change of resolution [Mah,12]. In this work, the HEVC standard is used as a compression attack for evaluating the performance of the proposed scheme.
2.5 Backup in Cloud System Cloud computing is a paradigm where a huge pool of systems are connected in private or public computing networks, to provide dynamically scalable infrastructure for the application of data and file storage. With the advent of this technology the cost of computation, application hosting, content storage and delivery is reduced significantly. It has the potential to transform a data center from a capital-intensive set up to a variable priced environment and 29
Chapter 2
Background of Watermarking based Video Coding
cloud computing is a practical approach to experience direct cost benefits. The idea of cloud computing depends on a very fundamental principal of reusability of Information Technology capabilities [Aro, 12]. The difference between cloud computing brings and traditional concepts of grid computing, distributed computing, utility computing and autonomic computing is to broaden horizons across organizational boundaries. The privacy and the security of personal information is extremely important. The personal information is being turned over to another organization, it is vital to ensure that the information is safe and the people who need to access it are able to do. The risk of personal information sent to a cloud provider might be kept indefinitely or used for other purposes. That information could also be accessed by government agencies, domestic or foreign [Bar, 11]. For the TV channels, there is a need of backup of all the media files which transfer through different communication medium. These files contain a huge amount of data when they are taken from the live streaming for both (audio and video). HEVC standard can reduce the file size and the watermarked video store in a one free cloud system i.e., Drop box, One Drive, Mozy, Google Drive, ZipCloud, etc. in a compression form.
30
Chapter 3
Proposed Scheme Design and Implementation
Chapter Three Proposed Scheme Design and Implementation 3.1 Introduction In this work, a robust watermarking scheme for high resolution video files is proposed. The robustness test is done by using one of the most recent coding algorithms for the high compression ratio, which is a HEVC standard. In this chapter, the design and the implementation steps are presented through the flowchart and the algorithm’s pseudo code. The video data could be either online or offline based on the test strategies. The input video is first extracted into frames and then converted from RGB color space to the of YUV space. The motion detection algorithm for the change detection is applied to separate the intra from inters frames. Two types of watermark are used, video profile and logo of the satellite TV, for embedding into intra and inter frames respectively. The video file and logo is converted into a bit stream then a bit stream is embedded into the LSB of the pixel of each of YUV. Finally, the watermarked video is objectively evaluated based on PSNR and Normalized Correlation. The HEVC coding system is used as an attacker tool for compressing the video files. Beside the brief introduction, this chapter gives the description of the video test sequences with their specifications in section (3.2). And then, in section (3.3) the structure of the algorithm design and implementation is illustrated by the general framework layout. In section (3.4 and 3.5) the flowchart or the algorithm steps of each part, embedding and extraction, process of the proposed scheme is presented. Finally the attacker tool for the compression is described in section (3.6). The general structure of the proposed system is shown in figure (3.1). 32
Chapter 3
Proposed Scheme Design and Implementation
Source video Online or Offline
Split into frames
Change detection (Normalized Correlation)
Intra frame
Inter frames
Color conversion RGB2YUV
Color conversion RGB2YUV
Add text
Add logo
Watermarked intra frame
Watermarked inter frames
Color conversion YUV2RGB All frames convert to video Apply attack by HEVC Extraction Stage Evaluate the Robustness (Normalized Correlation)
Fig. (3.1) General Structure of the proposed system 33
Chapter 3
Proposed Scheme Design and Implementation
3.2 General Framework Layout The general framework of the system consists of different phases starting by the change detection and ending up by evaluating the performance of the proposed method. 3.2.1 Change Detection (Intra and Inter Frame Separation) After taking the extracted frames form the input video files, a method based on normalized correlation is used to detect the objects in movement in successive frames. This leads to separate intra from inter frames. The flowchart of the change detection algorithm for both intra and inter frames is shown in figure (3.5 and 3.6) respectively. Begin
Load Video Frame
Counter=0 Set Max Threshold Calculate NC
No
If (NC ≥ Max)
Yes Mark as intra Counter ++ A
NC: Normalized Correlation
Fig. (3.2) flowchart of Intra frame depiction 34
Chapter 3
Proposed Scheme Design and Implementation Begin
Load Video Frames
Counter=0 Set Min Threshold Calculate NC
No
If (NC ≤ Min)
Yes Mark as inter Counter ++ B
NC: Normalized Correlation
Fig. (3.3) flowchart of Inter frames depiction
Figure (3.4) presents the test result output of the 1K video sequence using normalized correlation measurement for 1000 frames.
Frames from 1 to 1000
Fig. (3.4) Change detection for 1000 frames 35
Chapter 3
Proposed Scheme Design and Implementation
3.2.2 Color Space Conversion Color space is a scheme by which one can specify, create and visualize color. Color is defined by its attributes of brightness, hue and colorfulness. According to the computer system color is the amounts of red, green and blue. A color is thus usually specified using three co-ordinates, or parameters. These parameters describe the position of the color within the color space being used. They do not tell us what the color is, which depends on what color space is being used. RGB is an additive color system based on tri-chromatic method. It is easy to implement but non–linear with visual perception. It is a device which is dependent a specification of colors is semi–intuitive. RGB is very common, being used in virtually every computer system as well as television, video etc. Figure (3.5) depicts the component of digital image (RGB).
Fig. (3.5) the composition of digital image (RGB and YUV) 36
Chapter 3
Proposed Scheme Design and Implementation
YUV is the television transmission color spaces, sometimes known as transmission primaries. These color spaces separates RGB into luminance and chrominance information and are useful in compression applications (both digital and analogue). This space is a device dependent but is intended for use under strictly defined conditions within closed systems. Figure (3.6) shows the sample of color conversion between RGB and YUV.
b) YUV frame
a) RGB frame
Fig. (3.6) RGB color conversion to YUV
In figure (3.7) the pseudo code of the color conversion is given. Color Conversion Algorithm Input: A, R, G, B for image1 Input: A, R, G, B for image2 Output: Y, U, V Bitmap image1; Bitmap image2; Calculate Y, U, V Y = ((66 * R + 129 * G + 25 * B + 128) >> 8) + 16; U = ((-38 * R - 74 * G + 112 * B + 128) >> 8) + 128; V = ((112 * R - 94 * G - 18 * B + 128) >> 8) + 128;
Fig. (3.7) Color conversion pseudo code [Msd, 15] 37
Chapter 3
Proposed Scheme Design and Implementation
3.3 Watermark Embedding Process In the proposed approach, the embedded watermark must be invisible to human eyes and robust enough to some frame processing operations. To meet these requirements, a bit of binary pixel value (0 or 1) which is logo or text is embedded (additive method) to a selected pixel value of the host frames. Before insertion, the host frames color system (RGB) is converted to another color space (YUV) and then the color pixel values are sorted in the ascending or descending way to find out the low and high pixel values in the host frames. In the embedding process, the watermark is added not directly to the original pixel values of YUV component but to the selected pixel values based on the sort of algorithm calculation in LSB plain of YUV channel. Following formula is used for embedding watermark.
Where,
is watermarked image,
is original image,
is watermark and
is constant (α) and (0 < α < 1). Figure (3.11) presents the flowchart of the embedding process in spatial domain. The watermark used is logo and text. In change detection step the Intra and Inter frames are separated and the watermark added is as follows: Adding binary text as a watermark from video profile to those pixel values in intra frame that have been sorted according to the minimum or maximum value between (0-255). And then the position and the pixel values are saved into the Text file as a key for the extraction process. Adding binary watermark as (logo image) into the inter frames according to the size of watermark. The watermark bits are inserted based on the number of detected inter frames from the change detection step. And then the position and the pixel values are saved 38
Chapter 3
Proposed Scheme Design and Implementation
into the Text file as a key for the extraction process. Figure (3.8 and 3.9), show the flowchart of the watermark embedding process.
Begin
A
Get first intra frame Color conversion RGB2YUV
Video profile
Load watermark bits Calculate PSNR
Sort intra frame pixel values - Ascending set as Min - Descending set as Max Add watermark into LSB of Min pixel value Color conversion YUV2RGB Watermarked Frame
End
Fig. (3.8) Watermark embedding process flowchart for intra frame 39
Chapter 3
Proposed Scheme Design and Implementation
Begin
B Get inter frames
Color conversion RGB2YUV
Logo
Find W size divided by No of Frames (Threshold) Calculate PSNR
Sort inter frame pixel values
- Ascending set as Min - Descending set as Max
If No of Frames = Threshold
No
Yes
Binary Watermark
Add Watermark
Color conversion YUV2RGB
Watermarked Frames
End
Fig. (3.9) Watermark embedding process flowchart for inter frames 40
Chapter 3
Proposed Scheme Design and Implementation
3.4 Watermark Extraction Process In the watermark extraction process, the original and the watermarked frames are required in order to extract the watermark. The objective of the watermark extraction process is to reliably obtain an estimate of the original watermark from a possibly distorted version of the watermarked frames. The detection process requires knowledge of the watermark (the size of watermark for example (30x30) and the position of the watermark to add) and the key position. Once the watermark is extracted, one can compare it with the original watermark to determine the similarity between them. Finally, for the performance test of the watermarking scheme, the effect of the involved parameters when the watermarked image subjected to the HEVC attack will be explained in the next chapter. In figure (3.10) the flowchart of watermark extraction process of the given scheme is depicted.
41
Chapter 3
Proposed Scheme Design and Implementation
Begin Get watermarked intra Frame Color conversion RGB2YUV Read location and pixel values from Text file (Key)
Find diff. diff=WF-OF
If (diff=0)
No
Diff pixel value
Yes Extracted bit value
Get Text
End
Fig. (3.10) Watermark extraction process flowchart for intra frame
42
Chapter 3
Proposed Scheme Design and Implementation
Begin Watermarked Frames Color conversion RGB2YUV Read location and pixel values from Text file (Key)
Find diff. diff=WF-OF
Yes
End of Frames
No
Extracted watermark
Find Correlation between OW and EW
End
Fig. (3.11) Watermark extraction process flowchart for inter frames
43
Chapter 3
Proposed Scheme Design and Implementation
3.5 HEVC as Attacker Tool To develop a high robust watermarking scheme, some kinds of attack must be applied on the watermarked video to assess the performance of the proposed system in terms of imperceptibility and resistance against attacks. In this work, the x.265 (is the GUI or a tool for HEVC standard) program is used to test the robustness of the video watermarking. Figure (3.12) illustrates the robustness of the watermarked video using HEVC standard for compression purpose.
Load Frames
Add Watermark
In Spatial Domain
Watermarked Video
HEVC Attack
Fig. (3.12) Compression attack using HEVC
44
Chapter 3
Proposed Scheme Design and Implementation
In x.265 program the Quantization Parameter (QP) is used to change the rate control of the video. The range of QP value is taken from (0 – 51). Table (3.1) shows the effect of QPs on the following video: Table 3.1: QP effect on the video size
Video Property
QP
Size of Video
0
848 MB
5
411 MB
1K Video
10
202 MB
47.1 MB Size
20
52.8 MB
1280x720
26
25.7 MB
24 f / s
30
17.5 MB
00:03:47 Length
40
8.92 MB
45
7.3 MB
51
6.36 MB
45
Chapter 4
Test Results Evaluation
Chapter Four Test Results Evaluation 4.1 Introduction In this chapter, the test results are presented and discussed to reflect the performance capabilities of the established robust video watermarking scheme against HEVC as a compression attack. The proposed scheme is tested in both “embedding/extracting process before and after performing attacks” for a semiblind technique in a spatial domain. The conducted test is directed by tuning the involved parameters in embedding and extraction process, such adaptation of the parameter values are those which led to the best performance (i.e., imperceptibility of the watermark and high robustness). The performance of the system is evaluated by using the fidelity measure (PSNR) and the normalized correlation. The proposed method is implemented using (C#) programming language and is tested on SonyPC (equipped with CoreTMi5 CPU and 2.50 GHZ processor and 4.0 GB installed memory). In section (4.3), involved parameters that affect the performance of the proposed scheme are presented. First the change detection algorithm is adapted to separate the intra frame from inter frames and then the effect of the watermark size and the scale factor is mentioned. The watermark embedding and extraction process test results for both intra and inter frames are shown in section (4.4 and 4.5) respectively. All tests are performed on different high quality video samples (1K, 2K and 4K). The watermarked video is then subjected to the compression attack using HEVC standard for different values of the quantization parameters as explained in section (4.6). Finally, the performance of the proposed scheme is evaluated in terms of the objective quality (PSNR and NC) measurement.
47
Chapter 4
Test Results Evaluation
4.2 Test Sequences The implementation of the proposed watermarking scheme will be carried out on the YUV video sequences, which have different resolutions and framerates, covering the most use cases as possible as shown in tables (4.1-4.3): 4.2.1 Video Sample – 1K Table 4.1: Test sequence (HoneyBees)
Property
Value
Length
00:03:47
Frame Width
1280
Frame Height
720
Data Rate
1547 kbps
Total Bitrate
1739 kbps
Frame Rate
24 frames/second
Figure (4.1) shows one frame of the test sequence HoneyBees [1K].
Fig. (4.1) HoneyBees (1280x720) 48
Chapter 4
Test Results Evaluation
4.2.2 Video Sample – 2K
Table 4.2: Test sequence (BigBuckBunny)
Property
Value
Length
00:03:47
Frame Width
1920
Frame Height
1080
Data Rate
2996 kbps
Total Bitrate
3475 kbps
Frame Rate
30 frames/second
Figure (4.2) is the frame of the test sequence BigBuckBunny [2K]
Fig. (4.2) BigBuckBunny (1920x1080)
49
Chapter 4
Test Results Evaluation
4.2.3 Video Sample – 4K
Table 4.3: Test sequence (HoneyBees)
Property
Value
Length
00:03:47
Frame Width
4096
Frame Height
2304
Data Rate
22127 kbps
Total Bitrate
22320 kbps
Frame Rate
24 frames/second
Figure (4.3) is the frame of the test sequence HoneyBees [4K].
Fig. (4.3) HoneyBees (4096x2304)
50
Chapter 4
Test Results Evaluation
4.3 Involved Parameters In this section, the effect of the parameters that is involved during the development of the proposed scheme is described.
4.3.1 Change Detection Figure (4.4) shows the similarity value between successive frames in the 4K video sequence for 1000 frames. This test is done for the separation of the intra from the inter frames using two threshold (min and max) pixel values. Similarity
Frames from 1 to 1000
Fig. (4.4) Similarity versus frame number
For each video sample, 1000 frames are taken for the test, and the similarity is found between them by using two threshold values (min and max). According to the threshold values, the frames are separated into intra and inter frames. Table 4.4: Effect of threshold for separate Intra from Inter frames Threshold
Number of Inter frames
0.1
1
0.2, 0.3, 0.4, 0.5 and 0.6
3
0.7 and 0.8
5
0.9
8
Else all the other frames denoted as intra frames for any value of max threshold. 51
Chapter 4
Test Results Evaluation
4.3.2 Watermark Size and Scale Factor The watermark capacity may affect the performance of the proposed scheme. Figure (4.5) shows different watermark sizes for a logo types.
a)30x30
b)50x50
c)60x60 d)200x200
Fig. (4.5) Different watermark size
As an example, the effect of the logo size is presented in table (4.5). In all cases, the watermark size does not considerably affect the results when the size is about 60x60 but for the bigger size 200x200 the quality of the video is dramatically distorted. Table 4.5: Effect of watermark size Video Sample
Logo Size
PSNR/dB
30 x 30
53.02
4K
50 x 50
44.123
4096x2304
60 x 60
43.1377
200 x 200
38.66
The scale factor does not greatly change the PSNR and NC values as described below in details.
52
Chapter 4
Test Results Evaluation
4.4 Watermark Embedding Process In the following subsections, the results of the embedding process are presented for both minimum and maximum pixel values. 4.4.1 Video Sample – 1K A. Add Watermark to Maximum Pixel Values
Figure (4.6) illustrates the watermark embedding in a high pixel value for a video sample (1K).
a)Original Frame
b) Watermarked Frame
Fig. (4.6) 1K – Embedding watermark in high pixel value B. Add Watermark to Minimum Pixel Values
Figure (4.7) illustrates the watermark embedding in a low pixel value for a video sample (1K).
a)Original Frame
b) Watermarked Frame
Fig. (4.7) 1K – Embedding watermark in low pixel value
53
Chapter 4
Test Results Evaluation
4.4.2 Video Sample – 2K A. Add Watermark to Maximum Pixel Values
Figure (4.8) depicts the watermark embedding in a high pixel value for a video sample (2K).
a)Original Frame
b) Watermarked Frame
Fig. (4.8) 2K – Embedding watermark in high pixel value
B. Add Watermark to Minimum Pixel Values
Figure (4.9) depicts the watermark embedding in a low pixel value for a video sample (2K).
a)Original Frame
b) Watermarked Frame
Fig. (4.9) 2K – Embedding watermark in low pixel value
54
Chapter 4
Test Results Evaluation
4.4.3 Video Sample – 4K A. Add Watermark to Maximum Pixel Value
Figure (4.10) shows the watermark embedding in a high pixel value for a video sample (4K).
a)Original Frame
b) Watermarked Frame
Fig. (4.10) 4K – Embedding watermark in high pixel value
B. Add Watermark to Minimum Pixel Value
Figure (4.11) shows the watermark embedding in a low pixel value for a video sample (4K).
a)Original Frame
b) Watermarked Frame
Fig. (4.11) 4K – Embedding watermark in low pixel value 55
Chapter 4
Test Results Evaluation
4.5 Watermark Extraction Process The watermark extraction is the process of retrieving watermark position in watermarked frames. Different tests are performed for different video sequences (1K, 2K and 4K). In the following subsections, test results of the extraction process from both minimum and maximum pixel values for (1K, 2K and 4K) video sample are presented. 4.5.1 Video Sample – 1K A. Extract Watermark from Maximum Pixel Values Table (4.6) presents the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.6: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
0.99609 0.99609 0.99609 0.99619 0.99609
In figure (4.12) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.12) Extracted watermark – 1K
Table (4.7) presents the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. 56
Chapter 4
Test Results Evaluation Table 4.7: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
0.99699 0.99699 0.99699 0.99709 0.99699
In figure (4.13) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.13) Extracted watermark – 1K
Table (4.8) shows the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors. Table 4.8: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
0.996496 0.996496 0.996496 0.996657 0.996496
In figure (4.14) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
Fig. (4.14) Extracted watermark – 1K 57
e) Scale 1
Chapter 4
Test Results Evaluation
Table (4.9) shows the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors. Table 4.9: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
0.9976 0.9976 0.9976 0.997585 0.9976
In figure (4.15) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.5
c) Scale 1
Fig. (4.15) Extracted watermark – 1K
B. Extract Watermark from Minimum Pixel Values Table (4.10) illustrates the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.10: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
1 1 1 1 1
58
Chapter 4
Test Results Evaluation
In figure (4.16) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.16) Extracted watermark – 1K
Table (4.11) illustrates the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. Table 4.11: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.17) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.17) Extracted watermark – 1K
Table (4.12) illustrates the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors.
59
Chapter 4
Test Results Evaluation Table 4.12: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.18) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.18) Extracted watermark – 1K
Table (4.13) illustrates the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors. Table 4.13: NC versus scale factor Video Size
1280 x 720
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.19) the extracted watermark is shown for different scale factors.
60
Chapter 4
Test Results Evaluation
b) Scale 0.5
a) Scale 0.1
c) Scale 1
Fig. (4.19) Extracted watermark – 1K
4.5.2 Video Sample – 2K A. Extract Watermark from Maximum Pixel Values Table (4.14) presents the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.14: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
0.99749 0.99749 0.99749 0.997 0.99749
In figure (4.20) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
Fig. (4.20) Extracted watermark – 2K
61
e) Scale 1
Chapter 4
Test Results Evaluation
Table (4.15) presents the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. Table 4.15: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
0.997667 0.997667 0.997667 0.99763 0.997667
In figure (4.21) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.21) Extracted watermark – 2K
Table (4.16) shows the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors. Table 4.16: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
0.99736 0.99736 0.99736 0.99732 0.99736
In figure (4.22) the extracted watermark is shown for different scale factors.
62
Chapter 4
Test Results Evaluation
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.22) Extracted watermark – 2K
Table (4.17) shows the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors. Table 4.17: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
0.99749 0.99749 0.99749 0.997 0.99749
In figure (4.23) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.5
c) Scale 1
Fig. (4.23) Extracted watermark – 2K
63
Chapter 4
Test Results Evaluation
B. Extract Watermark from Minimum Pixel Values Table (4.18) illustrates the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.18: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.24) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.24) Extracted watermark – 2K
Table (4.19) illustrates the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. Table 4.19: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
1 1 1 1 1
64
Chapter 4
Test Results Evaluation
In figure (4.25) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.25) Extracted watermark – 2K
Table (4.20) illustrates the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors. Table 4.20: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.26) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.26) Extracted watermark – 2K
Table (4.21) illustrates the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors.
65
Chapter 4
Test Results Evaluation Table 4.21: NC versus scale factor Video Size
1920 x 1080
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.27) the extracted watermark is shown for different scale factors.
b) Scale 0.5
a) Scale 0.1
c) Scale 1
Fig. (4.27) Extracted watermark – 2K
4.5.3 Video Sample – 4K A. Extract Watermark from Maximum Pixel Values Table (4.22) presents the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.22: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
0.9973 0.9973 0.9973 0.99718 0.9973
In figure (4.28) the extracted watermark is shown for different scale factors. 66
Chapter 4
Test Results Evaluation
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.28) Extracted watermark – 4K
Table (4.23) presents the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. Table 4.23: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
0.99736 0.99736 0.99736 0.99732 0.99736
In figure (4.29) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.29) Extracted watermark – 4K
Table (4.24) presents the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors.
67
Chapter 4
Test Results Evaluation Table 4.24: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
0.997667 0.997667 0.997667 0.99763 0.997667
In figure (4.30) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.30) Extracted watermark – 4K
Table (4.25) shows the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors. Table 4.25: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
0.998487 0.998487 0.998487 0.99828 0.998487
In figure (4.31) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.5
c) Scale 1
Fig. (4.31) Extracted watermark – 4K
68
Chapter 4
Test Results Evaluation
B. Extract Watermark from Minimum Pixel Values Table (4.26) illustrates the normalized correlation between the original and the extracted watermark for a logo size (30x30) and different scale factors. Table 4.26: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
30 x 30
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.32) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.32) Extracted watermark – 4K
Table (4.27) illustrates the normalized correlation between the original and the extracted watermark for a logo size (50x50) and different scale factors. Table 4.27: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
50 x 50
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.33) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
Fig. (4.33) Extracted watermark – 4K 69
e) Scale 1
Chapter 4
Test Results Evaluation
Table (4.28) illustrates the normalized correlation between the original and the extracted watermark for a logo size (60x60) and different scale factors. Table 4.28: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
60 x 60
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.34) the extracted watermark is shown for different scale factors.
a) Scale 0.1
b) Scale 0.2
c) Scale 0.5
d) Scale 0.8
e) Scale 1
Fig. (4.34) Extracted watermark – 4K
Table (4.29) illustrates the normalized correlation between the original and the extracted watermark for a logo size (200x200) and different scale factors. Table 4.29: NC versus scale factor Video Size
4096 x 2304
Logo Size
Scale
NC
200 x 200
0.1 0.2 0.5 0.8 1
1 1 1 1 1
In figure (4.35) the extracted watermark is shown for different scale factors.
70
Chapter 4
Test Results Evaluation
b) Scale 0.5
a) Scale 0.1
c) Scale 1
Fig. (4.35) Extracted watermark – 4K
4.5.4 Embedding and Extracting Watermark from Intra Frame A. Embedding Process
In the embedding process, the profile of a company, TV channel for example, is hidden in the intra frame. At first, the text file is converted into the binary form, and then the bits are inserted to the LSB of pixel values according to the ascending and descending sort algorithms as shown in figure (4.36). Company Name: NRT TV
0100001101101111011011010111000 0011000010110111001111001001000 0001001110011000010110110101100 1010011101000100000010011100101 0010010101000010000001010100010 1011000001010010101100110100101 1001000110010101101111001000000 1010100011010010111010001101100 0110010100111010001000000100010 0011011110110001101110101011011 b) Binary
Video Title: Documentary Date: 1/1/2016 Code: 0001 About Video: [Video] Sykes-Picot reaches 100-year anniversary as Kurds push towards independence a) Text
Fig. (4.36) Text to binary conversion B. Extraction Process
In the extraction process, the watermarked intra frame is subtracted from the original intra frame and compares the result with the key file that created from embedding process. The binary data is then extracted from the watermarked frame and converted again to the text file. In all cases before and after attack the similarity is always equal to one (i.e., the original text is similar to the extracted text).
71
Chapter 4
Test Results Evaluation
4.6 Watermarked Video Attack In this section, the effect of the compression attack on the proposed scheme in terms of imperceptibility and robustness is described. 4.6.1 Video Sample – 1K A. Maximum Pixel Value
Case 1: logo size 30x30
Table (4.30) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.30: NC for different QP – logo size 30x30
Video Size
1280 x 720
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.6606 0.6605 0.6605 0.6582 0.6605
0.6309 0.6309 0.6309 0.62969 0.6309
0.56737 0.5664 0.5664 0.56588 0.5664
In figure (4.37), the extracted watermark is shown for different QP and scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.37) Extracted watermark after attack for (30x30) logo size
72
Chapter 4
Test Results Evaluation Case 2: logo size 50x50
Table (4.31) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.31: NC for different QP – logo size 50x50
Video Size
1280 x 720
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.749866 0.748759 0.748759 0.746158 0.748759
0.744169 0.742272 0.742272 0.7396 0.742272
0.672 0.668495 0.668495 0.666625 0.668495
In figure (4.38) the extracted watermark is shown for different QP and scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.38) Extracted watermark after attack for (50x50) logo size
73
Chapter 4
Test Results Evaluation Case 3: logo size 60x60
Table (4.32) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors. Table 4.32: NC for different QP – logo size 60x60
Video Size
1280 x 720
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.7563 0.75498 0.75498 0.7522 0.75498
0.7438 0.742 0.742 0.739999 0.742
0.6884 0.6858 0.6858 0.683896 0.6858
In figure (4.39), the extracted watermark is shown for different QP and scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.39) Extracted watermark after attack for (60x60) logo size
74
Chapter 4
Test Results Evaluation Case 4: logo size 200x200
Table (4.33) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors. Table 4.33: NC for different QP – logo size 200x200
Video Size
1280 x 720
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.7788 0.7777 0.7777 0.7757 0.7777
0.7736 0.77245 0.77245 0.77566 0.77245
0.7442 0.7415 0.7415 0.73958 0.7415
In figure (4.40), the extracted watermark is shown for different QP versus a scale factor. QP
α 10
20
0.1
0.2
75
30
Chapter 4
Test Results Evaluation
0.5
0.8
1
Fig. (4.40) Extracted watermark after attack for (200x200) logo size
B. Minimum Pixel Value
Case 1: logo size 30x30
Table (4.34) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.34: NC for different QP – logo size 30x30
Video Size
1280 x 720
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.915 0.9149 0.9149 0.91545 0.9149
0.91017 0.909257 0.909257 0.909589 0.909257
0.78005 0.7794 0.7794 0.7785 0.7794
76
Chapter 4
Test Results Evaluation
In figure (4.41) the extracted watermark is shown for different QP and scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8 1
Fig. (4.41) Extracted watermark after attack for (30x30) logo size
Case 2: logo size 50x50
Table (4.35) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.35: NC for different QP – logo size 50x50
Video Size
1280 x 720
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.93556 0.93509 0.93509 0.935796 0.93509
0.9339 0.93325 0.93325 0.9336 0.93325
0.8506 0.850195 0.850195 0.8504786 0.850195
In figure (4.42) the extracted watermark is shown for different QP versus a scale factor.
77
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.42) Extracted watermark after attack for (50x50) logo size
Case 3: logo size 60x60
Table (4.36) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors. Table 4.36: NC for different QP – logo size 60x60
Video Size
1280 x 720
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.949579 0.949287 0.949287 0.9496896 0.949287
0.941239 0.940736 0.940736 0.941117 0.940736
0.8567 0.85636 0.85636 0.856258 0.85636
In figure (4.43), the extracted watermark is shown for different QP versus a scale factor.
78
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.43) Extracted watermark after attack for (60x60) logo size
Case 4: logo size 200x200
Table (4.37) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors. Table 4.37: NC for different QP – logo size 200x200
Video Size
1280 x 720
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.98239 0.98193 0.98193 0.98215 0.98193
0.981 0.9803 0.9803 0.98145 0.9803
0.95769 0.9561 0.9561 0.9563 0.9561
In figure (4.44), the extracted watermark is shown for different QP versus a scale factor. 79
Chapter 4
Test Results Evaluation QP
α 10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.44) Extracted watermark after attack for (200x200) logo size 80
Chapter 4
Test Results Evaluation
4.6.2 Video Sample – 2K A. Maximum Pixel Value
Case 1: logo size 30x30
Table (4.38) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.38: NC for different QP – logo size 30x30
Video Size
1920 x 1080
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.3989 0.3989 0.3989 0.38989 0.3989
0.376 0.376 0.376 0.3698 0.376
In figure (4.45) the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.45) Extracted watermark after attack for (30x30) logo size
81
Chapter 4
Test Results Evaluation Case 2: logo size 50x50
Table (4.39) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.39: NC for different QP – logo size 50x50
Video Size
1920 x 1080
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.6008 0.60089 0.6009 0.6021 0.6009
0.556 0.556 0.556 0.552 0.556
0.4986 0.4986 0.4986 0.49056 0.4986
In figure (4.46), the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.46) Extracted watermark after attack for (50x50) logo size
Case 3: logo size 60x60
Table (4.40) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors.
82
Chapter 4
Test Results Evaluation Table 4.40: NC for different QP – logo size 60x60
Video Size
1920 x 1080
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.63678 0.63678 0.63678 0.630346 0.63678
0.60999 0.60999 0.60999 0.60765 0.60999
0.58256 0.58256 0.58256 0.57987 0.58256
In figure (4.47), the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.47) Extracted watermark after attack for (60x60) logo size
Case 4: logo size 200x200
Table (4.41) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors.
83
Chapter 4
Test Results Evaluation Table 4.41: NC for different QP – logo size 200x200
Video Size
1920 x 1080
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.78457 0.78457 0.78457 0.7806 0.78457
0.73612 0.73612 0.73612 0.73612 0.73612
0.6898 0.6898 0.6898 0.68457 0.6898
In figure (4.48), the extracted watermark is shown for different QP versus a scale factor.
QP α 10
20
0.1
0.2
0.5
84
30
Chapter 4
Test Results Evaluation
0.8
1
Fig. (4.48) Extracted watermark after attack for (200x200) logo size
B. Minimum Pixel Value
Case 1: logo size 30x30
Table (4.42) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.42: NC for different QP – logo size 30x30
Video Size
1920 x 1080
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.6636 0.664 0.664 0.66599 0.664
0.63915 0.651656 0.65656 0.6528 0.651656
0.5719 0.590779 0.59078 0.591307 0.59078
In figure (4.49), the extracted watermark is shown for different QP versus a scale factor.
85
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.49) Extracted watermark after attack for (30x30) logo size
Case 2: logo size 50x50
Table (4.43) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.43: NC for different QP – logo size 50x50
Video Size
1920 x 1080
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.69676 0.69676 0.69676 0.69034 0.69676
0.66566 0.66566 0.66566 0.66098 0.66566
0.59767 0.59767 0.59767 0.590469 0.59767
In figure (4.50), the extracted watermark is shown for different QP versus a scale factor.
86
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.50) Extracted watermark after attack for (50x50) logo size
Case 3: logo size 60x60
Table (4.44) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors. Table 4.44: NC for different QP – logo size 60x60
Video Size
1920 x 1080
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.724896 0.724896 0.724896 0.7208 0.724896
0.70987 0.70987 0.70987 0.70543 0.70987
0.68478 0.68478 0.68478 0.68 0.68478
In figure (4.51), the extracted watermark is shown for different QP versus a scale factor.
87
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.51) Extracted watermark after attack for (60x60) logo size
Case 4: logo size 200x200
Table (4.45) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors. Table 4.45: NC for different QP – logo size 200x200
Video Size
1920 x 1080
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.87069 0.8624 0.8624 0.8621 0.8623
0.8457 0.8395 0.8395 0.8395 0.83939
0.7885 0.78475 0.78475 0.7838 0.7838
In figure (4.52), the extracted watermark is shown for different QP versus a scale factor. 88
Chapter 4
Test Results Evaluation QP
α 10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.52) Extracted watermark after attack for (200x200) logo size 89
Chapter 4
Test Results Evaluation
4.6.3 Video Sample – 4K A. Maximum Pixel Value
Case 1: logo size 30x30
Table (4.46) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.46: NC for different QP – logo size 30x30
Video Size
4096 x 2304
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.3622 0.3622 0.3622 0.3574 0.3622
0.3443 0.3443 0.3443 0.3392 0.3443
0.3258 0.3258 0.3258 0.31588 0.3258
In figure (4.53), the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.53) Extracted watermark after attack for (30x30) logo size
90
Chapter 4
Test Results Evaluation Case 2: logo size 50x50
Table (4.47) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.47: NC for different QP – logo size 50x50
Video Size
4096 x 2304
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.556 0.556 0.556 0.552 0.556
0.48046 0.48046 0.48046 0.48435 0.48046
0.3957 0.3957 0.3957 0.392667 0.3957
In figure (4.54), the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.54) Extracted watermark after attack for (50x50) logo size
Case 3: logo size 60x60
Table (4.48) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors.
91
Chapter 4
Test Results Evaluation Table 4.48: NC for different QP – logo size 60x60
Video Size
4096 x 2304
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.58597 0.58597 0.58597 0.5808 0.58597
0.53488 0.53488 0.53488 0.53066 0.53488
0.4758 0.4758 0.4758 0.46987 0.4758
In figure (4.55), the extracted watermark is shown for different QP versus a scale factor. QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.55) Extracted watermark after attack for (60x60) logo size
Case 4: logo size 200x200
Table (4.49) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors.
92
Chapter 4
Test Results Evaluation Table 4.49: NC for different QP – logo size 200x200
Video Size
4096 x 2304
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.64359 0.64347 0.64347 0.6463 0.64347
0.6121 0.6121 0.6121 0.60469 0.6121
0.48096 0.48096 0.48096 0.47223 0.4808
In figure (4.56), the extracted watermark is shown for different QP versus a scale factor. α
QP 10
20
0.1
0.2
0.5
93
30
Chapter 4
Test Results Evaluation
0.8
1
Fig. (4.56) Extracted watermark after attack for (200x200) logo size
B. Minimum Pixel Value
Case 1: logo size 30x30
Table (4.50) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (30x30) and different scale factors. Table 4.50: NC for different QP – logo size 30x30
Video Size
4096 x 2304
Logo Size
Scale
30 x 30
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.4447 0.4447 0.4447 0.44389 0.4447
0.4346 0.4346 0.4346 0.42986 0.4346
0.3908 0.3908 0.3908 0.39079 0.3908
In figure (4.57), the extracted watermark is shown for different QP versus a scale factor.
94
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.57) Extracted watermark after attack for (30x30) logo size
Case 2: logo size 50x50
Table (4.51) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (50x50) and different scale factors. Table 4.51: NC for different QP – logo size 50x50
Video Size
4096 x 2304
Logo Size
Scale
50 x 50
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.5708 0.60089 0.6009 0.6021 0.6009
0.5219 0.519399 0.519399 0.51786 0.519399
0.494 0.489 0.489 0.4876 0.489
In figure (4.58), the extracted watermark is shown for different QP versus a scale factor.
95
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.58) Extracted watermark after attack for (50x50) logo size
Case 3: logo size 60x60
Table (4.52) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (60x60) and different scale factors. Table 4.52: NC for different QP – logo size 60x60
Video Size
4096 x 2304
Logo Size
Scale
60 x 60
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.64999 0.644695 0.644695 0.644549 0.644695
0.61688 0.6444 0.6444 0.64544 0.6444
0.5625 0.57035 0.57035 0.569757 0.57035
In figure (4.59), the extracted watermark is shown for different QP versus a scale factor.
96
Chapter 4
Test Results Evaluation QP
α
10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.59) Extracted watermark after attack for (60x60) logo size
Case 4: logo size 200x200
Table (4.53) illustrates the NC between the original and extracted logo regarding to the effect of QP for a logo size (200x200) and different scale factors. Table 4.53: NC for different QP – logo size 200x200
Video Size
4096 x 2304
Logo Size
Scale
200 x 200
0.1 0.2 0.5 0.8 1
NC
NC
NC
QP 10
QP 20
QP 30
0.84467 0.84467 0.84467 0.841678 0.84467
0.81198 0.81198 0.81198 0.8089 0.81198
0.76877 0.76877 0.76877 0.763698 0.76877
In figure (4.60), the extracted watermark is shown for different QP versus a scale factor.
97
Chapter 4
Test Results Evaluation
α
QP 10
20
30
0.1
0.2
0.5
0.8
1
Fig. (4.60) Extracted watermark after attack for (200x200) logo size
98
Chapter 4
Test Results Evaluation
4.7 Performance Evaluation The performance of the proposed scheme before and after attack is explained in detail in the next sections. 4.7.1 Video Sample – 1K A. Maximum Pixel Values
Table (4.54) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the highest pixel values. Table 4.54: PSNR for different QP and logo size PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
50.465
38.677
38.465
37.416
50 x 50
46.6465
38.25
38.06
37.1328
60 x 60
44.8
37.96
37.8
36.96
200 x 200
34.527
28.72
28.68
28.358
Figure (4.61) depicts the effects of logo size and QP on PSNR values. 60 50
PSNR
40 30x30
30
50x50
20
60x60
10
200x200
0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.61) Effect of logo size and QP on PSNR
99
Chapter 4
Test Results Evaluation
B. Minimum Pixel Value
Table (4.55) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the lowest pixel values. Table 4.55: PSNR for different QP and logo size
PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
55.12
38.82
38.6
37.53
50 x 50
54.888
38.818
38.59
37.53
60 x 60
54.71
38.81
38.596
37.53
200 x 200
38.43
33.83
33.79
33.387
Figure (4.62) depicts the effects of logo size and QP on PSNR values.
60 50
PSNR
40 30x30
30
50x50
20
60x60 200x200
10 0
Without
QP10
QP20
QP30
Without and QP
Fig. (4.62) Effect of logo size and QP on PSNR
100
Chapter 4
Test Results Evaluation
4.7.2 Video Sample – 2K A. Maximum Pixel Value
Table (4.56) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the highest pixel values. Table 4.56: PSNR for different QP and logo size PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
44.89
37.658
36.38
34.026
50 x 50
41.6
34.4497
33.8
33.756
60 x 60
39.67
33.07579
33.038
32.984
200 x 200
30.72
28.38
28.026
27.939
Figure (4.63) depicts the effects of logo size and QP on PSNR values. 50 45 40
PSNR
35
30
30x30
25
50x50
20
60x60
15
200x200
10
5 0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.63) Effect of logo size and QP on PSNR
101
Chapter 4
Test Results Evaluation
B. Minimum Pixel Value
Table (4.57) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the lowest pixel values. Table 4.57: PSNR for different QP and logo size PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
47.2675
34.5967
34.7187
34.4
50 x 50
44.123
33.869
34.2587
34.1795
60 x 60
43.1377
34.89
34.9966
34.399
200 x 200
35.16
32.87
32.496
31.93
Figure (4.64) depicts the effects of logo size and QP on PSNR values. 50 45 40
PSNR
35 30
30x30
25
50x50
20
60x60
15
200x200
10 5
0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.64) Effect of logo size and QP on PSNR
102
Chapter 4
Test Results Evaluation
4.7.3 Video Sample – 4K A. Maximum Pixel Value
Table (4.58) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the highest pixel values. Table 4.58: PSNR for different QP and logo size PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
52.82
37.54
36.154
34.4
50 x 50
41.6
34.4497
33.8
33.756
60 x 60
39.67
33.07579
33.038
32.984
200 x 200
30.72
28.38
28.026
27.939
Figure (4.65) depicts the effects of logo size and QP on PSNR values. 60 50
PSNR
40 30x30
30
50x50
20
60x60 200x200
10 0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.65) Effect of logo size and QP on PSNR
103
Chapter 4
Test Results Evaluation
B. Minimum Pixel Value
Table (4.59) shows the PSNR values between the original and the watermarked video before and after attack for different logo size. This evaluation is taken when the watermark is added into the lowest pixel values. Table 4.59: PSNR for different QP and logo size PSNR/ dB Logo Size
QP
Without Attack
10
20
30
30 x 30
53.02
37.5
36.85
34.39
50 x 50
44.123
35.869
34.2587
34.1795
60 x 60
43.1377
34.89
34.0966
33.399
200 x 200
38.66
30.263
29.445
28.87
Figure (4.66) depicts the effects of logo size and QP on PSNR values. 60 50
PSNR
40 30x30
30
50x50
20
60x60 200x200
10 0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.66) Effect of logo size and QP on PSNR
104
Chapter 4
Test Results Evaluation
Table (4.60) illustrates that 4K video sample for logo size 50x50 is better than 1K and 2K in case without attack, but after attack 1K video sample is getting better. The similarity is calculated when the watermark inserted into maximum pixel values. Table 4.60: NC versus video samples – high pixel value NC Logo size
Video Sample
QP Without QP
50 x 50
10
20
30
1K
0.99699
0.749866
0.744169
0.672
2K
0.997667
0.5708
0.556
0.4986
4K
0.998569
0.556
0.48046
0.3957
Table (4.61) illustrates that in all video sample for logo size 50x50, the watermark is completely extracted (NC = 100%). However, after attack 1K video sample is getting better. The similarity is calculated when the watermark is inserted into minimum pixel values. Table 4.61: NC versus video samples – low pixel value NC Logo size
Video Sample
QP Without QP
50 x 50
10
20
30
1K
1
0.93556
0.9339
0.8506
2K
1
0.69676
0.66566
0.59767
4K
1
0.5708
0.5219
0.494
105
Chapter 4
Test Results Evaluation
Figure (4.67) shows the PSNR between the original and the watermarked video for different video sequences. The PSNR is calculated when the watermark is inserted into maximum pixel values. 50
45 40 35 PSNR
30 25
1K
20
2K
15
4K
10 5
0 Without
QP10
QP20
QP30
Without and QP
Fig. (4.67) PSNR before and after attack - high pixel value
Figure (4.68) shows the PSNR between the original and the watermarked video for different video sequences. The PSNR is calculated when the
PSNR
watermark is inserted into minimum pixel values. 55 50 45 40 35 30 25 20 15 10 5 0
1K 2K
4K
Without
QP10 QP20 Without and QP
QP30
Fig. (4.68) PSNR before and after attack - low pixel value
106
Chapter 4
Test Results Evaluation
To compare the above results when the watermark is embedded in both minimum and maximum pixel values, the similarity between the original and the extracted watermark is presented in figure (4.69, 4.70 and 4.71) respectively when the logo size is 50x50 applied on the 1K, 2K and 4K video sequences. .
NC
1K Video 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Max Min
Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.69) NC values for 1K video
NC
2K Video 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Max Min
Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.70) NC values for 2K video
107
Chapter 4
Test Results Evaluation
4K Video 1 0.9
0.8 0.7
NC
0.6 0.5 0.4
Max
0.3
Min
0.2 0.1 0 Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.71) NC values for 4K video
The PSNR is also calculated between the original and the watermarked video and is presented in figure (4.72, 4.73 and 4.74) respectively when the logo size is 50x50 applied on the 1K, 2K and 4K video sequences. 1K Video 60 50
PSNR
40 Max
30
Min
20 10 0 Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.72) PSNR values for 1K video
108
Chapter 4
Test Results Evaluation
2K Video 60
50
PSNR
40 Max
30
Min 20 10
0 Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.73) PSNR values for 2K video
4K Video 60 50
PSNR
40 Max
30
Min
20 10 0
Without
QP10
QP20
QP30
Without Attack and QP
Fig. (4.74) PSNR values for 4K video
109
Chapter 4
Test Results Evaluation
In all cases, the imperceptibility and the robustness are optimal when the watermark is embedded in the lowest pixel values before performing attack. While after applying the HEVC attack for different values of quantization parameters, the NC and PSNR values are close to each other.
110
Chapter 5
Conclusions and Future Works
Chapter Five Conclusions and Future Works 5.1 Conclusions In this thesis, the embedding and the extraction processes for high quality digital video watermarking are studied and some improvements are proposed to enhance its performance. In the previous chapters, the concepts of watermarking system design and implementation are discussed for the spatial domain (semi-blind). The robustness of the proposed scheme, which is based on the similarity measure between original and extracted watermark, is tested for different video quality (1K, 2K and 4K). In this chapter, a list of remarks derived from the investigation of the test results of chapter four is presented. Also, some suggestions for future works are presented; which may enhance the overall system performance. 1. The test results led to better performance and a high robustness when the watermark is embedded in the minimum pixel values (see table 4.32 and 4.33). 2. Test results show that the imperceptibility and the robustness are optimal when the watermark is embedded in the lowest pixel values before performing attack (see figure 4.45, 4.46 and 4.47). 3. The proposed scheme shows that by applying the HEVC attack for different values of quantization parameters, the NC and PSNR values are varied as follows: a. When QP=10, the PSNR = 38.25 dB and NC= 0.7498 for the video sample 1K.
112
Chapter 5
Conclusions and Future Works
b. When QP=20, the PSNR = 38.06 dB and NC= 0.7441 for the video sample 1K. c. When QP=30, the PSNR = 37.13 dB and NC= 0.672 for the video sample 1K. According to the obtained results (see figures 4.46 and 4.47), the NC and PSNR values are decreased for higher video quality (2K and 4K). 4. The experimental results show that when the watermark size is small (30x30), the performance is getting better in terms of PSNR values (see table 4.26). 5. The experimental results show that when the watermark size is small, the similarity between original and the extracted watermark is low (see table 4.23). In contrary, larger watermark size yields better correlation (see 4.25).
5.2 Suggestions for Future Works During the discussion of the test results, the following suggestions could be taken into consideration for future works: 1. For further performance improvement, other video qualities (8K and 3D video) can be used for testing. 2. This research can be extended by implementing the proposed scheme in the frequency domain and comparing it to the spatial domain. 3. Developing and implementing the embedding and extraction processes in the source code of HEVC standard. 4. Adding another level of security to the watermark before embedding into the video file using different cryptography algorithms. 5. A subjective quality assessment could be used in order to compare it with the objective one, which is used in this work PSNR.
113
Chapter 5
Conclusions and Future Works
6. Implement the embedding process in the median pixel value and compare with maximum and minimum pixel values. 7. Add watermark to Y, U and V channel separately and then compare with YUV channel.
114
References [Ais, 14] Aisheng Yang, Huanqiang Zeng, King Ngi Ngan, Miaohui Wang, Canhui Cai, “Efficient Perceptual Rate Control Algorithm for High Efficiency Video Coding”, Vol. MTAP-D, No. 00946, 2014.
[Ali, 16] Ali A.Elrowayati, M.F.L. Abdullah, Azizah Abd Manaf, Abd. S. Alfagi “Robust HEVC Video Watermarking Scheme Based on Repetition-BCH Syndrome Code”, International Journal of Software Engineering and Its Applications Vol. 10, No. 1, pp. 263-270, 2016.
[Alk, 14] Alka N. Potkar, Saniya M. Ansari “Review on Digital Video Watermarking Techniques”, International Journal of Computer Applications (0975 – 8887), Volume 106, No.11, November 2014.
[Ami, 13] Amit Singh, Susheel Jain, Anurag Jain “A Survay: Digital Video Watermarking”, International Journal of Scientific & Engineering Research, Volume 4, Issue 7, July2013, ISSN 2229-5518.
[Aro, 12] R. Arokia Paul Rajan, S. Shanmugapriyaa “Evolution of Cloud Storage as Cloud Computing Infrastructure Service”, IOSR Journal of Computer Engineering (IOSRJCE), ISSN: 2278-0661, Volume 1, Issue 1 (May-June 2012).
115
[Asi, 15] Asim Naveed, Yasir Saleem, Nisar Ahmed, Aasia Rafiq “Performance Evaluation and Watermark Security Assessment of Digital Watermarking Techniques”, Sci.Int. (Lahore), 27(2), 1271-1276, 2015, ISSN 1013-5316; CODEN: SINTE 8.
[Bar, 11] Barrie Sosinsky “Cloud Computing Bible”, Published by Wiley Publishing, Inc., Indianapolis, Indiana Published simultaneously in Canada ISBN: 978-0-470-90356-8 , 2011.
[Bog, 15] “http://www.bogotobogo.com/VideoStreaming/h265_hevc.php”Accessed Nov 2015.
[Gar, 12] Gary J. Sullivan , Jens-Rainer Ohm , Woo-Jin Han , and Thomas Wiegand , “Overview of the High Efficiency Video Coding (HEVC) Standard” IEEE Transactions on Circuits and Systems for Video Technology, Vol. 22, No. 12, December 2012.
[Ger, 00] Gerrit Cornelis Langelaar “Real-time Watermarking Techniques for Compressed Video Data”, Thesis Delft University of Technology - With ref. - With Summary in Dutch Printed by Universal Press, Veenendaal, ISBN 90-9013190-6, 2000.
116
[Gop, 13] Gopika V Mane, G. G. Chiddarwar, “ Review Paper on Video Watermarking Techniques”, International Journal of Scientific and Research Publications, Volume 3, Issue 4, April 2013, ISSN 2250-3153.
[Hsu, 15] Hsueh-Ming Hang, Wen-Hsiao Peng, Chia-Hsin Chan and Chun-Chi Chen, “Towards the Next Video Standard: High Efficiency Video Coding”, This work was supported in part by the NSC, Taiwan under Grants 98-2622-8-009-011 and 982219-E-009-015, December 2010.
[Jam, 09] Jamal Hussein and Aree Mohammed “Robust Video Watermarking using MultiBand Wavelet Transform”, IJCSI International Journal of Computer Science Issues, Vol. 6, No. 1, 2009, ISSN (Online): 1694-0784, ISSN (Print): 1694-0814.
[Jan, 12] Jani Lainema, Frank Bossen, Woo-Jin Han, Junghye Min and Kemal Ugur, “Intra Coding of the HEVC Standard”, IEEE Transactions on Circuits and Systems for Video Technology, VOL. 22, No. 12, December 2012.
[Jen, 12] Jens-Rainer Ohm, Gary J. Sullivan, Heiko Schwarz, Thiow Keng Tan, and Thomas Wiegand, “Comparison of the Coding Efficiency of Video Coding Standards— Including High Efficiency Video Coding (HEVC)”, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 22, No. 12, December 2012.
117
[Jon, 15] Jong-Hyeok Lee, Kyung-Soon Jang, Byung-GyuKim, Seyoon Jeong, and Jin Soo Choi “Fast Video Encoding Algorithm for the Internet of Things Environment Based on High Efficiency Video Coding”, Hindawi Publishing Corporation, International Journal of Distributed Sensor Networks, Volume 2015, Article ID 146067.
[Kar, 14] Karuna Gubbi Shivashankar Sastri “Complexity Reduction in Inter Layer Inter Prediction in Scalable High Efficiency Video Coding”, Master of Science in Electrical Engineering, the University of TEXAS at ARLINGTON, August 2014.
[Kum, 06] A. Kumer “Improved Content Based Image Watermarking”, MSc thesis at University of Louisiana State, 2006.
[Mah, 12] Mahsa T. Pourazad, Colin Doutre, Maryam Azimi, and Panos Nasiopoulos “HEVC: The New Gold Standard for Video Compression”, IEEE Consumer Electronics Magazine, July 2012, 2162-2248.
[Man, 10] Manish K Thakur, Vikas Saxena ; J. P. Gupta “A performance analysis of objective video quality metrics for digital video watermarking”, Computer Science and Information Technology (ICCSIT), 2010 3rd IEEE International Conference, Volume: 4
118
[Mia, 15] Miaohui Wang, KingNgiNgan, and Hongliang Li, “An Efficient Frame-Content Based Intra Frame Rate Control for High Efficiency Video Coding”, IEEE Signal Processing Letters, Vol. 22, No. 7, July 2015.
[Msd, 15] https://msdn.microsoft.com/en-us/library/windows/desktop/dd206750(v=vs.85).aspx, Accessed May 2015.
[Nit, 13] Nitin A. Shelke, Dr. P.N.Chatur, “A Survey on Various Digital Video Watermarking Schemes” International Journal of Computer Science & Engineering Technology (IJCSET), ISSN : 2229-3345 Vol. 4 No. 12, Dec 2013.
[Pre, 14] Preeti Parashar, Rajeev Kumar Singh “A Survey: Digital Image Watermarking Techniques”, International Journal of Signal Processing, Image Processing and Pattern Recognition Vol. 7, No. 6, 2014, pp. 111-124.
[Pri, 13] Priyanka Premnath, Srilatha.L.Rao, Kendaganna Swamy, “A Review Paper on Forensic Watermarking”, International Journal of Research in Advent Technology, Volume 1, Issue 5, December 2013, E-ISSN: 2321–9637.
[Rad, 11] Radu Ovidiu Preda, Nicolae Vizireanu “New Robust Watermarking Scheme for Video Copyright Protection in the Spatial Domain”, U.P.B. Sci. Bull., Series C, Vol. 73, Iss. 1, 2011, ISSN 1454-234x. 119
[Rad, 12] Radu Ovidiu Preda, Cristina Oprea, Ionuţ Pirnog, Lucian Andrei Perişoară “Robust Digital Video Watermarking in the Spatial and Wavelet Domain”, the Seventh International Conference on Digital Telecommunications, IARIA, 2012, ISBN: 9781-61208-193-9.
[Raj, 13] Rajkumar Buyya, Christian Vecchiola, S Thamarai Selvi, “Mastering Cloud Computing”, Published by Tata McGraw Hill Education Private Limited, 7 West Patel Nagar, New Delhi 110 008, Typeset at BeSpoke Integrated Solutions, Puducherry, India 605 008 and printed at Magic International Pvt. Ltd., Plot No. 26E, Sector-31(INDUSTRIAL), Site-IV, Greater Noida – June 2013.
[Rin, 11] Rini T. Paul, “Review of Robust Video Watermarking Techniques”, IJCA Special Issue on “Computational Science - New Dimensions & Perspectives”NCCSE, 2011.
[Sal, 14] Salahuddin Swati, Khizar Hayat, Zafar Shahid, “A Watermarking Scheme for High
Efficiency
Video
Coding
(HEVC)”,
PLOS
ONE,
10.1371/journal.pone.0105613, August 2014.
[See, 14] Dr.V.Seenivasagam, S.Subbulakshmi and S.Radhamani “A Survey on Video Watermarking and Its Applications”, Computing Terminologies and Research Development Conference Held at SCAD College of Engineering and Technology, India, Volume 4, Issue 3, March 2014, ISSN: 2277 128X.
120
[Suj, 15] C. N. Sujatha, P. Satyanarayana, “An Improved Hybrid Color Image Watermarking under Various Attacks”, International Journal of Advanced Research in Computer and Communication Engineering Vol. 4, Issue 3, March 2015.
[Vco, 15] “http://www.vcodex.com/hevc-an-introduction-to-high-efficiency-coding/”Accessed Oct. 3, 2015.
[Vip, 11] Dr. Vipula Singh “Digital Watermarking: A Tutorial”, Cyber Journals: Multidisciplinary Journals in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT), January Edition, 2011.
[Viv, 14] Vivienne Sze, Madhukar Budagavi, Gary J. Sullivan “High Efficiency Video Coding (HEVC) Algorithms and Architectures”, Springer International Publishing Switzerland 2014, ISBN 978-3-319-06895-4 (eBook).
[Yim, 13] Yimin Zhou “Intra frame constant rate control scheme for high efficiency video coding”, IEEE, Computing, Networking and Communications (ICNC), Jan. 2013, ISBN:978-1-4673-5287-1.
121
الخالصت الفیدیوىات عالیو الجودة یحتاج الى مستوى امنیة لتدفق المعمومات بین االجيزة المختمفو (کومبیوتر ،تمفزیونات ،االقمار الصناعی ،االیباد ،النوتباد و اليواتف النقالة) وعبر احد قنوات االتصال. ان في ىذا البحث ،ابتدع عالمة مائیة رقمیة قویة لمفیدیو معتمدا عمى تشفیر الفیدیو بكفائة عالیة .اليدف منو ىو تصمیم و تنفیذ مشروع لمفیدیو مع عالمة مائیة فى نطاق مكاني .و اضیفت مستوى امنیة ميمة عن طریق وضع عالمة مائیة قویة و التى تستخدم آلخفاء الممف الشخصي لممالك فى داخل الفیدیو .ىذه اآلمنیة یحافظ عمى حق النشر لمحتویات الفیدیو المرقم .و اخی ار النظام المقترح یدعم فایالت الفیدیو فى احد انظمة السحاب الكبیوتریة المشتركة مثل (زیب كالود ،كاربونایت ،دروب بوكس و موزى). یستخرج اآلطارات من فایل الفیدیو المدخل اوال و بعد ذلك یحول من نظام المون ( )RGBالى (.)YUV ان خوارزمیة اكتشاف الحركة لكشف التغییر ینفذ لفصل اآلطارات الداخمیة و البینیة .نوعان من العالمات المائیة استخدمت (الممف الشخصي لمفیدیو و شعار القنوات التمیفزیونیة) لتضمین اآلطارات الداخمیة و البینیة عمى التوالى .و ختاما ,الفیدیو ذات العالمة المائیة تقیم بموضوعیة اعتمادا عمى قمة اآلشارة الى نسبة الضوضاء و ارتباط التطبیع .ان نظام التشفیر( )HEVCاستخدمت كأداة مياجمة لضغط فایالت الفیدیو. النتائج یظير بأن مختمف الفیدیوىات من حیث الجودة ( 2K, 1kو )4Kحین یضمن العالمة المائیة الى الخالیا ذات قیم منخفضة فأن الكفائة یكون افضل مقارنة بالنتائج التى یحصل عمیيا فى الخالیا ذات القیم العالیة .ان حجم العالمة المائیة ایضا یؤثر عمى اداء النظام المقترح .ان قیمة ( )PSNRیكون عالیا حین یكون حجم العالمة المائیة صغیرا .ان تشابو ( )NCبین العالمة المائیة اآلصمیة و المستخرجة یكون عالیا حین یكون حجم العالمة المائیة كبی ار بعد الضغط بتطبیق مقیاس (.)HEVC
تشفیر و وضع عالمة مائیة للفیدیو بکفائة عالیة
رسالت مقذمت الً كليت الخجارة-جامعت السليماويت كجسء مه مخطلباث ويل درجت الماجسخير في حقىيت المعلىماث
مه قبل
وياز عسیس علً بكالىریىش في االحصاء والحاسباث جامعت السليماويت
بأشراف
د .ئاري علً دمحم أسخار مساعذ
1436
2016
پىخخە ڤیدیۆ وایاتەکان پێىیستً تە ئاستێکً خۆپارێسي هەیە تۆ گىاستىەوەي زاویاري لە ویَىان ئامێرە جیاوازەکان (کۆمپیىتەر ،تەلەفسیۆن ،ماوگً دەستکرد ،ئایپاد ،وۆتپاد و ئامێرە پەیىەودیە گەڕۆکەکان) وە لە ڕێً یەکێک لە کەواڵەکاوً پەیىەودي .لەم لێکۆڵیىەوەیەدا ،ویشاوەیەکً ئاوي دیجیتاڵً تەهێس داهێىراوە کە پشت تە شفرەکردوً ڤیدیۆ تە کاراییەکً تەرز دەتەسترێت .ئاماوج لێی داڕشته و جێثەجێکردوً سیستەمێک تۆ ڤیدیۆ لەگەڵ ویشاوەیەکً ئاوي لە میاوەیەکً شىێىیدا .وە ئاستیکً خۆپارێسي گروگً تۆ زیادکراوە لەڕێی داواوً ویشاوەیەکً ئاوي تەهێس کە تۆ شاردوەوەي فایلً کەسً خاوەوەکەي تەکاردێت لە ڤیدیۆکەدا .ئەم خۆپارێسیە پارێسگاري لە مافً تاڵوکردوەوەي واوەرۆکً ڤیدیۆ ژمارەییەکە دەکات.وە لە کۆتاییدا ئەم سیستمە پێشىیاز کراوە پاڵپشتً فایلً ڤیدیۆیً دەکات لە یەکێک لە سیستمەکاوً هەوري کۆمپیىتەري هاوتەش وەکى (زیة كالود ،كارتىوایت ،دروب تىكس و مىزي). لە سەرەتادا چىارچێىەکان دەردەهێىرێت لە فایلً ڤیدیۆکە دواتر لە سیستمً ڕەوگً ()RGBیەوە دەکرێت تە ( .)YUVئەلگۆریسمً دۆزیىەوەي جىوڵە تۆ دۆزیىەوەي گۆڕاوکاري جێثەجێدەکرێت تۆ جیاکردوەوەي چىارچێىە واوخۆییەکان لە وێىاویەکان .دوو جۆر لە ویشاوە ئاوییەکان تکارهێىراوە (فایلً کەسً ڤیدیۆکە و ئارمً کەواڵە تەلەفسیۆویەک)ە تۆ لەخۆگرتىً چىارچێىە واوخۆیی و وێىاویەکان دوا تە دواي یەک .وە لە کۆتاییدا ,ئەو ڤیدیۆیەي ویشاوەي ئاوي لەگەڵدایە تاتەتیاوە هەڵدەسەوگێىرێت تە پشت تەسته تە لىتکەي سیگىاڵەکەي تۆ ڕێژەي ژاوەژاو وە هاوپەیىەودي ئاساییکردوەوە .سیستمً تەشفرەکردوً ( )HEVCتەکارهێىراوە وەکى ئامرازي پەستاودوً فایلە ڤیدیۆییەکان. ئەوجامەکان دەریدەخەن کە ڤیدیۆ جیاوازەکان لە ڕووي چۆوێتیەوە ( )4k, 2k, 1kکاتێک کە ویشاوەي ئاوییان تۆ خاوە قیمەت وسمەکان زیاددەکرێت ئەوە لە ڕووي کاراییەوە تاشترە تەتەراورد لەگەڵ ئەو ئەوجاماوەي کە لە خاوە قیمەت تەرزەکاوەوە دەست دەکەون .هەروەها قەتارەي ویشاوە ئاوییەکە کاردەکاتە سەر ئاستً جێثەجێکردوً سیستمە پێشىیار کراوەکە .قیمەتً ( )PSNRتەرز دەتێت کاتێک قەتارەي ویشاوە ئاوییەکە تچىوک تێت .لێکچىوي ( )NCلەوێىان ویشاوە ئاوییە ڕەسەوەکە و هەڵهێىجراوەکە تەرز دەتێت کاتێک قەتارەي ویشاوە ئاوییەکە گەورە تێت دواي پەستاودن تە پێىەري (.)HEVC
بە شفرەکردن و دانانى نیشانەى ئاوى بۆ ڤیدیۆ بە کاراییەکى بەرز
وامەیەکە پێشکەشە بۆ ئەوجىمەوً کۆلێجً بازرگاوً-زاوکۆي سلێماوً وەک بەشێک لە پێذاویسخيەکاوً بەدەسخهێىاوً بڕواوامەي ماسخەري زاوسج لە حەکىەلۆجياي زاوياري
لە الیەن وياز عسیس علً
بە سەرپەرشخً
د .ئاري علً دمحم پرۆفيسۆري یاریذەدەر
2٧16
2016