IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

International Journal of Research in Information Technology (IJRIT) www.ijrit.com

ISSN 2001-5569

Edge Based Image Segmentation Techniques: A Survey Prabhjot Singh1, Er Deepak Sharma 2 1

M.Tech Final Year Student, Deptt. of ECE,MMEC, MMU, Mullana Ambala, Haryana, India [email protected] 2

Assitant Professor, Deptt of ECE, MMEC, MMU, Mullana Ambala, Haryana, India [email protected]

Abstract Image segmentation is the process of partitioning an image into multiple segments, so as to change the representation of an image into something that is more meaningful and easier to analyze. Several general-purpose algorithms and techniques have been developed for image segmentation. Images are considered as one of the most important medium of conveying information, in the field of computer vision, by understanding images the information extracted from them can be used for other tasks. In the computer vision domain and image analysis several important researches can be performed in the segmentation of video with dynamic background. Image segmentation is most of judging or analyzing function in image processing and analysis. Image segmentation refers to partition of an image into different regions that are homogenous or similar and inhomogeneous in some characteristics. This paper describes the image segmentation methods in the field of computer vision.

Keywords: Mean Shift, Watershed, Normalized Cuts, Graph Cuts, Edge Based Segmentation.

1. Introduction Image processing refers to processing of a 2D picture by a computer. An image is defined in the “real world” is considered to be a function of two real variables. An image is an array, or a matrix, of square pixels (picture elements) arranged in columns and rows, which identifies a point in the image and the corresponding matrix element value identifies the gray level at that point. Image processing is a method to convert an image into digital form and perform some operations on it, in order to get an enhanced image or to extract some useful information from its image domain [1].Image Segmentation is the process of partitioning a digital image into multiple regions or sets of pixels. Actually, partitions are different objects in image which have the same texture or color. The result of image segmentation is a set of regions that .All of the pixels collectively cover the entire image, or a set of contours extracted from the image in a region are similar with respect to some characteristic or computed property, such as color, intensity, or texture. Image segmentation is the first step and also one of the most difficult tasks of image analysis, which has objective of extracting information which is represented in the form of data from image via image segmentation, feature measurement and object representation. The result of segmentation of image is considerably depends on the accuracy of feature measurement. Image segmentation is the computer-aided so that the computerization of medical image segmentation plays an important role in medical imaging applications. Image segmentation process that subdivides an image into its constituent parts and extracts Prabhjot Singh, IJRIT

455

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

those parts of interest or objects. Automatic image segmentation also done but the most critical task is that the segmentation result affect all the subsequent processes of image analysis [4].Image (and video) segmentation is a critical step of image analysis occupying the middle layer of image engineering, which means it is influenced not only from data but also from human factors. Image analysis having the feature measurement, object representation and description, and even the higher level tasks such as object classification and scene interpretation [3].

2. Image Segmentation Methods 2.1 Mean shift method Is a non-parametric feature-space analysis technique, a so-called mode seeking algorithm. Mean shift is a procedure for locating the maxima of a density function given discrete data sampled from that function. It is useful for detecting the modes of this density. This is an iterative method, and we start with an initial estimate x. Let a kernel function
( − )be given. This function determines the weight of nearby points for re-estimation of the mean. Typically Gaussian kernel on the distance to the current estimate is 

used,
( − ) = ∥ 

∥ . The weighted mean of the density in the window determined by k is () =

∑ ∈() ( 

)  ∑

 ∈()

( 

)

(1)

Where ()is the neighborhood of x, a set of points for which k(x) ≠0. The MS algorithm is a robust feature-space analysis approach which can be applied to discontinuity preserving smoothing and image segmentation problems. It can significantly reduce the number of basic image entities, and due to the good discontinuity preserving filtering characteristic, the salient features of the overall image are retained. The latter property is particularly important in the partitioning of natural images, in which only several distinct regions are used in representing different scenes such as sky, lake, sand beach, person, and animal, where as other information within a region is often less important and can be neglected. However, it is difficult to partition a natural image into significative regions to represent distinct scenes, depending only on the MS segmentation algorithm. The main reason is that the MS algorithm is an unsupervised clustering-based segmentation method, where the number and the shape of the data cluster are unknown a priori. Moreover, the termination of the segmentation process is based on some region-merging strategy applied to the filtered image result, and the number of regions in the segmented image is mainly determined by the minimum number of pixels in a region, which is denoted as M (i.e., regions containing less than M pixels will be eliminated and merged into its neighboring region) [5].

2.2 Edge based segmentation In edge based segmentation technique boundary on an image or an edge is defined by the local pixel intensity gradient. An estimation of the first order derivative of the image function is called a gradient. The magnitude of the gradient for a given image f(x,y) can be calculated as  =   +  

(2)

The direction of gradient is represented as  = 2(, )

(3)

Here, gradients in directions x and y are expressed as Gx and Gy. Edge-based techniques are fast in computation and usually in this approach a priori information about image content is not required. The Prabhjot Singh, IJRIT

456

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

most general problem of this approach is that often the edges do not enclose the object completely. In this segmentation technique the direction and magnitude can be presented as images. A post processing step of linking or grouping edges is required to structure closed boundaries neighboring regions. Weighed summation of the pixel intensities in a small neighborhood can be represented as a numerical array in this method which is called as kernel/ window/mask. In a two 3X3 mask the following matrices are used. −1 0 +1 (4)  = −2 0 +2 −1 0 +1 +1  = 0 −1

+2 0 −2

+1 0 −1

(5)

To compute Gx and Gy, first and second mask are used respectively. Finally, joining Gx and Gy using the mentioned equation, gradient magnitude image is obtained [6].

2.3 Watershed technique A grey-level image may be seen as a topographic relief, where the grey level of a pixel is interpreted as its altitude in the relief. A drop of water falling on a topographic relief flows along a path to finally reach a local minimum. Intuitively, the watershed of a relief corresponds to the limits of the adjacent catchment basins of the drops of water. In image processing, different watershed lines may be computed. In graphs, some may be defined on the nodes, on the edges, or hybrid lines on both nodes and edges. Watersheds may also be defined in the continuous domain. There are also many different algorithms to compute watersheds. For a segmentation purpose, the gradient magnitude (i.e., the length of the gradient vectors) is interpreted as elevation information. Watershed transformation also called, as watershed method is a powerful mathematical morphological tool for the image segmentation. It is more popular in the fields like biomedical and medical image processing, and computer vision [7]. In geography, watershed means the ridge that divides areas drained by different river systems. If image is viewed as geological landscape, the watershed lines determine boundaries which separate image regions. The watershed transform computes catchment basins and ridgelines (also known as watershed lines), where catchment basins corresponding to image regions and ridgelines relating to region boundaries [8]. Segmentation by watershed embodies many of the concepts of the three techniques such as threshold based, edge based and region based segmentation.

3. Normalized Cuts The Ncut method can be considered as a classification method. In most image segmentation applications, the Ncut method is applied directly to the image pixels, which are typically of very large size and thus require huge computational complexity. For example, to use the Ncut method in a gray image has to be decimated into a size of 160× 160 pixels or smaller. In summary, it is difficult to get real-time segmentation using the Ncut method. In the proposed method, the Ncut method is applied to the segmented regions instead of the raw image pixels. As such, it eliminates the major problem of the Ncut method that requires prohibitively high complexity. By applying the Ncut method to the preprocessed regions rather than the raw image pixels, the proposed method achieves a significant reduction of the computational cost and, therefore, renders real-time image segmentation much more practically implemental. On the other hand, due to some approximation in the implementation of the Ncut method, the segmentation processing of a graph exploiting the lower dimensional region-based weight matrix also provides more precise and robust partitioning performance compared to that based on the pixel-based weight matrix. In the normalized cut framework, we also model the image into a graph. We model each pixel of the image as a node in the graph, and set an edge between two nodes if there are similarities between them. The Prabhjot Singh, IJRIT

457

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

normalized cut framework is composed of two steps: similarity measurement and normalized cut process. The first step should be combined with feature extraction. The purpose of this step is to compute the similarity between pixels and this value is set as the weight on the edge. In order to model all the similarities of an image, all pair of pixels will contain an edge, which means if an image contains N pixels, there will be totally (N-1)N /2 edges in the corresponding graph. This kind of graph is called “complete graph” and needs a large memory space. To simplify the problem, sometimes we set edges between two nodes only when their distance is smaller than a specific threshold. A graph G (V, E) can be partitioned into two disjoint sets, A, B, A ∪ B = V, A ∩ B = ∅ by simply removing edges connecting the two parts. The degree of dissimilarity between these two pieces can be computed as total weight of the edges that have been removed. In graph theoretic language, it is called the cut The normalized cut then could be defined as: *+(,, -) =

./(0,1) 2334(0,5)

+

./(0,1) 2334(1,5)

(6)

Cut (A, B) = Summation of edge weights associated with the cut. Assoc (A, V) = Summation of edge weights associated with all the points in A [8].

3.1 Min-cut/max-flow algorithm for graph cuts The energy minimization procedures for machine vision and automated image segmentation are explored by Boykov and Kolmogorov in [8]. Based on augmenting paths, the min-cut/max flow algorithm is presented with two reusable and non-overlapping search trees- S: from source s and T: from sink t. Tree S has the direction of non-saturation from parent node to children and tree T has non-saturation from children to parent node. There can be active or passive node in S or T based on outer border and internal border respectively. Free nodes are those who are not in S or T considering the conditions 6 ⊂ 8, 9 ∈ 6, : ⊂ 8,  ∈ :, 6 ∩ : = ∅. The graphical illustration of this new algorithm is shown in figure

Figure1. Example of the min-cut/max-flow algorithm in graph cuts segmentation [9]. Red nodes indicate search tree S, blue nodes represent search tree T and yellow line is for the path from the source s to sink t. Free nodes are represented as black circle while active node is by A and passive node is by P. The combination of minimum s-t cut and maximum flow optimizations is accomplished with three steps-growth, augmentation and adoption in the segmentation procedure [9].

4. Graph Cut Method Graph cuts method has attracted a lot of attention for image segmentation. The graph cuts method is one of minimizing energy functions elegantly expressed as MRF (Markov random field), and the energy function consists of two terms; the first term, called data term, is to globally capture the characteristics of an image in feature space, and the second term, called smoothness term, is to preserve spatial information in an image domain. Thus, the graph cuts-based approach can combine abovementioned two approaches, Prabhjot Singh, IJRIT

458

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

clustering in feature space and preserving spatial relationship in image domain. As applied in the field of computer vision, graph cuts can be employed to efficiently solve a wide variety of low-level computer vision problems (early vision), such as image smoothing, the stereo correspondence problem, and many other computer vision problems that can be formulated in terms of energy minimization. Such energy minimizations problems can be reduced to instances of the maximum flow problem in a graph (and thus, by the max-flow min-cut theorem, define a minimal cut of the graph). Under most formulations of such problems in computer vision, the minimum energy solution corresponds to the maximum a posteriori estimate of a solution. Although many computer vision algorithms involve cutting a graph (e.g., normalized cuts), the term "graph cuts" is applied specifically to those models which employ a max-flow/min-cut optimization (other graph cutting algorithms may be considered as graph partitioning algorithms)."Binary" problems (such as denoising a binary image) can be solved exactly using this approach; problems where pixels can be labeled with more than two different labels (such as stereo correspondence, or denoising of a grayscale image) cannot be solved exactly, but solutions produced are usually near the global optimum. The theory of graph cuts was first applied in computer vision in the paper by Greig, Porteous and Seheult of Durham University. In the Bayesian statistical context of smoothing noisy (or corrupted) images, they showed how the maximum a posteriori estimate of a binary image can be obtained exactly by maximizing the flow through an associated image network, involving the introduction of a source and sink [10].

Figure 2. graph cut for image segmentation (source and sink) [11]. Notations: • Image:  ∈ ;<, , -=> • Output: Segmentation (also called opacity) S ∈ ;R=A (soft segmentation). For hard segmentation S ∈ ; 0 for background , 1 for foreground/object to be detected = • Energy function: Q(, 6, R, S)where C is the color parameter and λ is the coherence parameter. • Q(, 6, R, S) = Q4T4U + Q4VWUWXW Optimization: The segmentation can be estimated as a global minimum over 6: Z[\] Q(, 6, R, S) [10].

5. Conclusion With the numerous amounts of image segmentation techniques presented in this paper, the selection of segmentation approach depends on what quality of segmentation is required. Further, it also depends on what scale of information is required. From the above study of the various segmentation methods, it is observed that the graph cut approach gives the better segmentation to any of the arbitrary shapes.

Prabhjot Singh, IJRIT

459

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

References [1] H.P. Narkhede, “Review of Image Segmentation Techniques” International Journal of Science and Modern Engineering (IJISME) ISSN: 2319-6386, Volume-1, Issue-8, July 2013 [2] Shraddha Tripathi, Krishna Kumar, B.K.Singh, R.P.Singh, “Image Segmentation: A Review”, International Journal of Computer Science and Management Research, Vol 1 Issue 4 November 2012. [3]Zhang, Y. J, An Overview of Image and Video Segmentation in the last 40 years, Proceedings of the 6th International Symposium on Signal Processing and Its Applications, pp. 144-151, 2001. [4] D.L. Pham, Ch. Xu and J.L. Princo, “a survey on current methods in medical image segmentation” Annual Review of Biomedical Engineering, vol. 2, 2000. [5]M. Sonka, V. Hlavac and R. Boyle, “Image processing, analysis, and machine vision”, Third edition, Thomson, USA, 2008. [6]Issac N. Bankman, “Handbook of medical image processing and analysis”, Second edition, Academic press, USA, 2008. [7] Md. Shakowat Zaman Sarker, Tan Wooi Haw and Rajasvaran Logeswaran, “Morphological based technique for image segmentation”, International Journal of Information Technology, Vol. 14, No. 1. [8] Manisha Bhagwat, R. K. Krishna and Vivek Pise, “Simplifed Watershed Transformation”, International Journal of Computer Science and Communication, Vol.1, No. 1., pp. 175-177, 2010. [8] Wenbing Tao, Hai Jin, Senior Member, IEEE, and Yimin Zhang, Senior Member, IEEE, “Color Image Segmentation Based on Mean Shift and Normalized Cuts” IEEE Transactions on Systems, Man, and Cybernetics—Part B: Cybernetics, VOL. 37, NO. 5, OCTOBER 2007. [9]Boykov, Y., Kolmogorov, V, “An experimental comparison of min-cut/max-flow algorithms for energy minimization in vision”, IEEE Trans, Pattern Anal. Machine Intell, 26, 1124–1137, 2004. [10] Anjin Park, Jungwhan Kim, Seungki Min, Sungju Yun, Keechul Jung, “Graph Cuts based Automatic Color Image Segmentation using Mean Shift Analysis” IEEE Digital Image Computing: Techniques and Applications, 2008. [11]Yuri Boykov and Vladimir Kolmogorov “An Experimental Comparison of Min-Cut/Max-Flow Algorithms for Energy Minimization in Vision” IEEE Transactions on PAMI, Vol. 26, No. 9, pp. 11241137, Sept. 2004. [12] Jianwen Mo, Chaoxuan Wang, Tong Zhang, Hua Yua, “Improved graph-based image segmentation based on Mean Shift 1” Communications and Networking in China (CHINACOM), 2013 8th International ICST Conference on Digital Object Identifier: 10.1109/ChinaCom.2013.6694680, pp: 685 – 689, 2013. [13] Wei-Ming Chen, Sheng-Hao Jhang, “Improving Graph Cuts algorithm to transform sequence of stereo image to depth map” The Journal of Systems and Software 86, pp: 198– 210, 2013. [14] Bahadir Karasulu. “An approach based on simulated annealing to optimize the performance of extraction of the flower region using mean-shift segmentation” Applied Soft Computing 13 pp: 4763–4785, 2013. [15] Dedic, R.Allili, Madjid, “A novel cooperative approach for cardiac PET image segmentation” Information Science, Signal Processing and their Applications (ISSPA), 2012 11th International Conference on Digital Object Identifier: 10.1109/ISSPA.2012.6310474 pp: 11 – 16,2012. [16] B. Padmapriya, T. Kesavamurthi, H. Wassim Ferose, “Edge Based Image Segmentation Technique for Detection and Estimation of the Bladder Wall Thickness” Procedia Engineering 30, pp: 828 – 835, 2012. [17] Ety Navon, Ofer Miller1, Amir Averbuch, “Color image segmentation based on adaptive local thresholds” Image and Vision Computing 23, pp: 69–85, 2012. [18]W. X. Kang, Q. Q. Yang, R. R. Liang,“The Comparative Research on Image Segmentation Algorithms”, IEEE Conference on ETCS, pp. 703-707, 2009. [19]Mei Yeen Choong, Wei Yeang Kow, Yit Kwong Chin, Lorita Angeline, Kenneth Tze Kin Teo, “Image Segmentation via Normalised Cuts and Clustering Algorithm”, 2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia. Prabhjot Singh, IJRIT

460

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 3, March 2014, Pg: 455- 461

[20] K. K. Singh, A. Singh, “A Study of Image Segmentation Algorithms for Different Types of Images”, International Journal of Computer Science Issues, Vol. 7, Issue 5, 2010 [21] P. Karch, I. Zolotova, “An Experimental Comparison of Modern Methods of Segmentation”, IEEE 8th International Symposium on SAMI, pp. 247-252, 2010. [22] Zengwei Ju, Jingli Zhou, Xian Wang, Qin Shu,“Image segmentation based on adaptive threshold edge detection and mean shift” Software Engineering and Service Science (ICSESS), 24th,IEEE International Conference on Digital Object Identifier: 10.1109/ICSESS.2013.6615330 pp: 385 – 388,2013.

Prabhjot Singh, IJRIT

461