How to reduce light leakage and clipping in local-dimming liquid-crystal displays Seong-Eun Kim (SID Student Member) Joo-Young An Jong-Ju Hong Tae Wook Lee Chang Gone Kim Woo-Jin Song

Abstract — In conventional LCDs, the backlight is set to maximum luminance regardless of the image. For dark scenes, this approach causes light leakage and power waste. Especially, light leakage in dark scenes degrades the contrast ratio of LCDs; to circumvent this problem, local-dimming systems have been proposed. In these systems, the LED backlight is divided into several local blocks and the backlight luminance of each local block is controlled individually, and pixel values are adjusted simultaneously according to the luminance profile of the dimmed backlight. In this paper, a method of determining the LED backlight luminance of each local block depending on the image is proposed; this method significantly improves the image quality of LCDs. First, we introduce methods of quantifying light-leakage at dark gray levels and clipping at bright gray levels. Then, the proposed method to determine the dimming duty, which controls the LED backlight luminance by compromising between these two measures, was derived. The proposed algorithm preserves the original image with little clipping distortion and effectively reduces light leakage. Keywords — Liquid-crystal display (LCD), backlight local dimming, contrast ratio, power consumption, light leakage, clipping. DOI # 10.1889/JSID17.12.1051

1

Introduction

The liquid-crystal display (LCD) is the most popular display device in the flat-panel-display marketplace due to its excellent performance. However, conventional LCDs have serious light leakage of the liquid-crystal (LC) because the LC cannot completely obstruct the backlight luminance when the LC is fully closed to display black pixels.1 In addition, the backlight of conventional LCDs is always at maximum luminance regardless of the image even when displaying a dark image. Consequently, dark regions of images do not look dark enough and true black cannot be realized. This problem causes a low contrast ratio and unnecessary power consumption.2 To reduce the light-leakage problem, two types of modifications have been proposed. One is to improve the structure of the LC to reduce the transmittance of the LC at low gray levels. However, using this method alone is not sufficient to increase contrast ratio. The other modification is to dim the backlight when displaying dark scenes. These dimming techniques can be categorized into three classes: global dimming, line dimming, and block dimming. First, global-dimming methods were proposed.3–8 These methods consider the image, then dim all backlight luminance uniformly by an appropriate factor k and enhance the transmittance of the LC by 1/k. However, because bright regions and dark regions occur together in the image, global dimming methods do not reduce light leakage effectively to preserve bright regions. As an alternative, several local (line or block) dimming methods have been proposed, which divide the backlight

into several local blocks, then dim the backlight individually in these blocks.9–16 When LCDs were first developed, line fluorescent lamps were generally used as LCD backlights, so line dimming methods were proposed.9,10 Recently, due to the development of light-emitting-diode (LED) backlights, block-dimming methods have received more attention. For local block dimming, the backlights are generally grouped into rectangular blocks and the image is divided into non-overlapping regions corresponding to the backlight blocks, shown in Fig. 1. Thus, the luminance of backlight blocks can be dimmed independently according to the local image.11–16 Therefore, because LED backlights are dimmed only in dark regions, the contrast ratio increases and power consumption decreases. In addition, pixel values are adjusted locally according to the backlight luminance, so pixel luminance is preserved. However, by dimming the backlight to reduce light leakage, the maximum luminance for the brightest pixel is decreased, which causes the clipping problem at bright gray levels after pixel compensation. Therefore, in backlight local dimming, it is important to determine the backlight luminance of each local block to improve the quality of LCDs when bright regions and the dark regions occur in the same block. In this paper, we propose a novel local LED-backlight block-dimming algorithm to determine the backlight luminance in each local block. This algorithm considers both light leakage and clipping of local block images. We first define a light-leakage measure and a clipping measure. The proposed algorithm determines backlight luminance of each

Expanded revised version of a paper presented at the 15th International Display Workshops held in Niigata, Japan, December 3–5, 2008. S-E. Kim, J-J. Hong, and W-J. Song are with Pohang University of Science and Technology, Division of Electrical and Computer Engineering, San 31, Hyoja-dong, Nam-gu, Pohang, Gyungbuk, 790-784, Korea; telephone +82-54-279-2789, fax –8686, e-mail: [email protected]. J-Y. An, Solid Technologies, Gyeonggi-do, Korea. T. W. Lee and C. G. Kim, LG Display, Gyeonggi-do, Korea. © Copyright 2009 Society for Information Display 1071-0922/09/1712-1051$1.00

Journal of the SID 17/12, 2009

1051

FIGURE 2 — Pixel luminance of the LCD panel.

Therefore, a method is required that achieves a trade-off between the Average and the Max methods. For that reason, histogram-based backlight-dimming methods9–11 were proposed. In Ref. 9, when the dimming duty d is determined, the inverse of the cumulative distribution function of the entire image is used; this approach considers the characteristics of the entire image. In Ref. 10, the histogram is used to determine the total error caused by dimming the backlight, and the value of d that minimizes the total error. In Ref. 11, the major gray levels, small bright objects, and noise pixels of each image block can be balanced when determining the LED-backlight luminance. In this paper, we define a light-leakage measure and a clipping measure for each local block based on the histogram. Unlike the conventional histogram-based dimming methods, the proposed backlight-dimming algorithm balances the two measures by considering the correlation of inter-blocks.

FIGURE 1 — The structures of the LED backlight unit and liquid-crystal cells for local block dimming.

local block using a decision rule that determines optimal backlight luminance by comparing these measures. The proposed algorithm reduces the light-leakage effectively in dark regions with little clipping in bright regions. This paper is organized as follows. Section 2 introduces previous backlight local-dimming methods and the characteristics of LCDs. Section 3 presents the proposed algorithm. Section 4 contains experimental results and Sec. 5 presents conclusions.

2 2.1

LCD characteristics

Generally, the luminance of a pixel is determined by the product of the LC transmittance and the backlight luminance, as shown in Fig. 2. Furthermore, the pixel luminance of LCDs should be set such that luminance of all gray levels meet the ideal gamma characteristics, which are represented by the dashed line in Fig. 3. The ideal pixel luminance Yideal(g) is defined as

Yideal (g) =

FG g IJ g , H 255 K

(1)

Background Previous work

The critical problem of local dimming is how to choose the luminance of each LED local block according to the image. The simplest method is the Average method,14 which determines the LED luminance based on the average gray level of each image block. This method has the disadvantage that it may decrease the brightness and lose information in bright regions. Another method is the Max method,17 which sets the LED luminance of each block to the maximal gray level of each image block. The Max method has the disadvantages that it is very sensitive to noise and that it wastes power; it also suffers from the light-leakage problem.

1052

2.2

FIGURE 3 — Light-leakage problem of LCDs; the dashed line is an ideal gamma curve and the solid line represents the light leakage.

Kim et al. / How to reduce light leakage and clipping in local-dimming LCDs

for high gray levels cannot fulfill the ideal gamma characteristics even when pixel compensation is executed, so Y(g′, d) above a threshold Yc(d) is clipped owing to a shortage of backlight luminance, as represented by the solid line in Fig. 5(b), where Yc(d) is given by

Yc (d) = Y(255, d) = Yreal (255) ¥

FIGURE 4 — The relationship between pixel luminance and dimming duty.

where g is a gray level and γ is an ideal target gamma. However, LCs cause light leakage at low gray levels, so that conventional LCDs fail to meet the ideal gamma characteristics at low gray levels, which is presented by the solid line in Fig. 3. Let us denote the real measured pixel luminance of LCDs as Yreal(g), which is given by Yreal(g) = k(g) + Yideal(g),

(2)

d . 255

(6)

In addition, although the backlight is dimmed, Y(g, d) still has light leakage at low gray levels compared with Yideal(g) because Y(0, d) = (d/255)k(0) is always larger than Yideal(0) = 0, when d ≠ 0 and k(0) ≠ 0. Dimming the backlight reduces light leakage, but causes clipping at high gray levels. Thus, determination of the appropriate value of d must compromise between its effects on light leakage and clipping.

3 3.1

Proposed algorithm Proposed dimming duty decision rule

The major goal of dimming the backlight is to reduce light leakage and thereby increase the contrast ratio. However, clipping should not degrade the quality of the image. For this reason, the algorithm should simultaneously consider

where k(g) is the amount of light leakage at low gray levels. In practice, k(g) is calculated by the difference between the measured Yreal(g) and the ideal target Yideal(g) as follows: k(g) = max[Yreal(g) – Yideal(g), 0],

(3)

where max(A, B) means that the larger one between A and B is taken. To reduce the light leakage, the backlight should be dimmed for blocks that have many low gray levels. However, as the backlight is dimmed, the pixel luminance is also reduced for all gray levels. For LED backlights, the backlight luminance is controlled by a dimming duty d, which is an integer between 0 (black) and 255 (white). Thus, the effect of d on Yreal(g) should be investigated in detail. To investigate the relationship between d and Yreal(g), an experiment was conducted in which all pixels were set to the same gray level (i.e., 127, 223, or 255), and d was varied from 0 to 255. Then, the pixel luminance was measured in the fixed block. In Fig. 4, the pixel luminance Y(g, d) after dimming backlight is experimentally given by

Y(g, d) = Yreal (g) ¥

d . 255

(4)

As a result, if backlight-dimming algorithms reduce d to eliminate light leakage, they also reduce Y(g,d) for each gray level, as shown in Fig. 5(a). To compensate for the reduced luminance, gray levels should be adjusted using pixel compensation.15 The gray level g is enhanced by the compensation factor C(d) as follows: g′ = C(d) ⋅ g,

(5)

where C(d) is obtained by solving Y(g′, d) = Yideal(g), which is generally constituted by a look-up table. However, Y(g, d)

FIGURE 5 — The results of dimming the backlight: (a) As a result of dimming the backlight, pixel luminance is reduced for all gray levels. (b) Some high gray levels cannot be compensated due to reduced luminance.

Journal of the SID 17/12, 2009

1053

artifacts caused by light leakage and clipping. To consider the effect of these artifacts, we define an error as the difference between the ideal target luminance and the pixel-compensated luminance after dimming. Generally, subjective brightness, which is the intensity as perceived by the human visual system, is a logarithmic function of light intensity.18 Thus, two distortion measures based on subjective brightness are necessary. Because the backlight is divided into i rows and j columns of blocks, these measures must be determined separately for each block. The inverse gamma function is applied when luminance is converted to subjective brightness, so the subjective brightness error ec(g, di,j) caused by clipping is given by

gg - eYc (di, j )jg , 0OPP , Q

LMb MN

e c (g, d i, j ) = max Yideal (g)

-1

-1

(7)

where di,j is the dimming duty of the block in row i and column j. The clipping measure should consider both the distortion area and the quantity of the subjective brightness error. Therefore, the clipping measure Dc(i, j) is given by 255

Dc (i, j) = Â e c (g, d i, j ) ◊ ni, j (g),

(8)

3.2

Determining the weight parameter

The critical problem of the dimming duty decision rule is how to choose the value of αi,j of each block according to the image. For a given αi,j, as a dark region of a block becomes wider, the dimming duty of the block decreases according to Eq. (11). However, when we consider the scene of a dark sky with some bright stars, if the backlight luminance is too small due to a broad dark region, then the bright stars may look dim. Therefore, if a very dark region includes bright objects, backlight luminance is increased to preserve the brightness of objects in this region. Generally, we assume that the bright objects in a dark region have the maximum pixel value in the block. Therefore, if maximum luminance of a block is bright and an average luminance of adjacent blocks is dark, then αi,j must be increased to prevent clipping and preserve details since the diffusion of light by backlight of adjacent blocks is insufficient. Otherwise, αi,j must be decreased to prevent light leakage and increase contrast ratio. By this logic, αi,j should be proportional to a ratio of an average pixel value of adjacent blocks to the maximal pixel value of a current block. Assuming that adjacent blocks are in the M × M macro block centering around the current (i, j)-th block, as shown in Fig. 6, then αi,j is given by

g=0

a i, j =

where ni, j(g) is the number of pixels that have gray level g in the (i, j)-th block. Also, the light-leakage measure Dl(i, j) is given by 255

Dl (i, j) = Â e l (g, d i, j ) ◊ ni, j (g),

(9)

g= 0

MAX(i, j) 1

(M -1) 2

Â

(M -1) 2

,

(12)

 APL(i - a, j - b)

M 2 a =-(M -1) 2 b=-(M -1) 2

where APL(i, j) is the average of all pixel values, MAX(i, j) is the maximum of all pixel values in the block in row i and column j and M is the size of the macro block.

where

O jg - bYideal (g)gg , 0PPQ .

LM MNe

-1

e l (g, d i, j ) = max Y(g, d i, j )

-1

(10)

Determination of the appropriate value of d must compromise between the conflicting effects of clipping and light leakage. Thus, d should be determined by comparing Dc(i, j) and Dl(i, j). If Dc(i, j) > Dl(i, j), then artifacts caused by clipping at high gray levels are recognized more easily than artifacts caused by the light leakage. In that case, d is increased to decrease Dc(i, j). On the other hand, if Dc(i, j) ≤ Dl(i, j), d is decreased to reduce Dl(i, j). However, the weights of Dc(i, j) and Dl(i, j) vary according to the characteristics of the image. As a result, the proposed dimming duty decision rule to search the optimal dimming duty di,jo iteratively is given by di,jo = di,j

if

Dl(i, j) = αi,j Dc(i, j),

Investigation on the proposed algorithm

In some cases, the proposed algorithm may require many iterations to find di,jo that satisfies Dl(i, j) = αi,jDc(i, j). The

(11)

where αi,j is a weight parameter for the block in row i and column j. αi,j controls the relative importance of clipping and light leakage according to the image. The value of d that satisfies Eq. (11) can reduce the light leakage, thus minimizing the number of significant artifacts at high gray levels.

1054

3.3

FIGURE 6 — An M × M macro block centering around the (i, j)-th block.

Kim et al. / How to reduce light leakage and clipping in local-dimming LCDs

FIGURE 8 — Flow diagram of the procedure used in the experiments. FIGURE 7 — An efficient method for finding the optimal dimming duty iteratively, which reduces the complexity of the proposed algorithm.

complexity of the proposed algorithm increases as the number of iterations increases. Thus, we present an effective method to find di,jo. Figure 7 shows the effective method. The proposed method is presented in six steps for the block in row i and column j as follows: Step 1: initialize dmax = 255, dmin = 0, and di,j = APL(i, j). Step 2: calculate αi,j. Step 3: calculate Dl(i, j) and Dc(i, j) using di,j. Step 4: if Dl(i, j) > αi,jDc(i, j), then dmax = di,j, dmin is unchanged, else dmin = di,j, dmax is unchanged. Step 5: update di,j = ⎡(dmax + dmin)/2⎤.

Step 6: repeat step 3 – step 5 until dmax – dmin = 1. where ⎡⋅⎤ represents rounding off to the nearest integer.

4

Experimental results

The proposed algorithm was implemented in a full-HD 47in. LCD TV equipped with white LED backlights, which are grouped into 16 columns of eight blocks. To verify the proposed algorithm’s performance, three methods are compared in the LCD panel, such as the Average method, the Max method, and the proposed method. After dimming the backlight using the three methods, the pixel compensation proposed by Ref. 16 was executed, as shown in Fig. 8. In

FIGURE 9 — The resulting images of (a) not using any dimming methods; (b) the Average method; (c) the Max method; (d) the proposed method.

Journal of the SID 17/12, 2009

1055

FIGURE 10 — Comparison of dimming duties determined by the Average method, the Max method, and the proposed method in (a) a bright region and (b) a dark region. The weight parameters are represented in parenthesis.

addition, the size of macro block M and the ideal gamma γ were set to 3 and 2.2, respectively. Figure 9 shows the experimental results to compare the light leakage and details in the dark region of the three methods. The light leakage can plainly be recognized when any backlight local dimming is not applied in Fig. 9(a). In Fig 9(b), the Average method can reduce light leakage effectively because of the reduced backlight luminance, but details in dark regions are almost lost. In Fig 9(c), the Max method retains the details, but does not effectively reduce light leakage that occurs due to the bright backlight. On the other hand, the proposed algorithm preserves most image details in dark regions and reduces light leakage significantly, as shown in Fig. 9(d). Figure 10 shows values of d, which were determined by the three methods in six blocks that form a part of the test image shown in Fig. 9. The values of d generated by the proposed method were lower than those of both the Max method and the Average method in most regions, but were sufficiently large to minimize clipping distortion when bright objects exist in dark regions. The power consumption of the proposed algorithm and the conventional algorithms are demonstrated in Fig. 11. As can be seen in Fig. 11, the power consumption rate of the proposed algorithm is almost half that of the Max method without loss in brightness. Because the power consumption rate depends on the image contents, the power consump-

tion rate of the proposed algorithm is between that of the Max method and that of the Average method.

5

Conclusions

We have presented a noble method of determining the dimming duty of each local block by using the proposed decision rule. The proposed decision rule searches the optimal dimming duty through the comparison between the light-leakage measure and the clipping measure. Two measures have been introduced to reflect the magnitude and occurrence of subjective brightness error. Compared with conventional methods, the proposed algorithm preserves the original image with little clipping distortion and reduces light leakage effectively, so the contrast ratio is improved. In addition, power consumption is reduced due to the reduction of dimming duty.

Acknowledgment This work was supported by LG Display, the IT R&D Program of MKE/MCST/IITA (2008-F-031-01, Development of Computational Photography Technologies for Image and Video Contents), the HY-SDR Research Center at Hanyang University under the ITRC Program of MKE, and the Brain Korea (BK) 21 Program funded by Ministry of Education, Science and Technology, Korea.

References

FIGURE 11 — The average power consumption rates deduced from average dimming duties determined by the Average method, the Max method, and the proposed method.

1056

1 P. Yeh and C. Gu, Optics of Liquid Crystal Displays (Wiley, New York, 1999), Chap. 5. 2 A. Konno et al., “RGB color systems for LED backlight in IPS-LCD TVs,” SID Symposium Digest 36, 1380–1383 (2005). 3 I. Hwang et al., “Image synchronized brightness control,” SID Symposium Digest 32, 492–493 (2001). 4 N. Raman and G. Hekstra, “Dynamic contrast enhancement of liquidcrystal displays with backlight modulation,” Intl. Conf. on Consumer Electronics Digest of Technical Papers, 197–198 (2005). 5 L. Kerofsky and S. Daly, “Brightness preservation for LCD backlight dimming,” J. Soc. Info. Display 14, No. 12, 1111–1118 (2006).

Kim et al. / How to reduce light leakage and clipping in local-dimming LCDs

6 J. Stessen and J. Mourik, “Algorithm for contrast reserve, backlight dimming, and backlight boosting on LCD,” SID Symposium Digest 37, 1249–1252 (2006). 7 C-F. Hsu et al., “Backlight power reduction and image contrast enhancement using adaptive dimming for global backlight applications,” SID Symposium Digest 39, 776–779 (2008). 8 S.-J. Lee et al., “A power reduction method for LCD backlight based on human visual characteristics,” Intl. Conf. on Consumer Electronics Digest of Technical Papers, 1–2 (2008). 9 E. Y. Oh et al., “IPS-mode dynamic LCD-TV realization with low black luminance and high contrast by adaptive dynamic image control technology,” J. Soc. Info. Display 13, No. 3, 215–219 (2005). 10 P. D. Greef and H. G. Hulze, “Adaptive dimming and boosting backlight for LCD-TV systems,” SID Symposium Digest 38, 1332–1335 (2007). 11 H. Peng et al., “High contrast LCD TV using active dynamic LED backlight,” SID Symposium Digest 38, 1336–1338 (2007). 12 T. Shiga et al., “Power savings and enhancement of gray-scale capability of LCD TVs with an adaptive dimming technique,” J. Soc. Info. Display 16, No. 2, 311–316 (2008). 13 D.-M. Yeo et al., “Smart algorithms for local dimming LED backlight,” SID Symposium Digest 39, 986–989 (2008). 14 W. Lee et al., “White LED backlight control for motion blur reduction and power minimization in large LCD TVs,” J. Soc. Info. Display 17, No. 1, 37–45 (2009). 15 F. Lin et al., “Inverse of mapping function (IMF) method for image quality enhancement of high dynamic range LCD TVs,” SID Symposium Digest 38, 1343–1346 (2007). 16 H. Chen et al., “Locally pixel-compensated backlight dimming on LED-backlit LCD TV,” J. Soc. Info. Display 15, No. 12, 981–988 (2007). 17 H. Seetzen et al., “High dynamic range display systems,” ACM Trans. Graphics 23, No. 3, 760–768 (2004). 18 R. C Gonzalez and R. E. Woods, Digital Image Processing, 2nd edn. (Prentice Hall, Upper Saddle River, New Jersey, 2002), Chap. 2. Seong-Eun Kim received his B.S. degree in electronic and electrical engineering from Pohang U n i ve r s i t y o f S c ie n c e an d Tech no logy (POSTECH), Korea, in 2004. Since 2004, he has been a Research Assistant at the Department of Electronic and Electrical Engineering, POSTECH, where he is currently working toward his Ph.D. degree. His research interests include multimedia signal processing, signal processing for display, and adaptive signal processing.

Joo-Young An received his B.S. and M.S. degrees in electronic and electrical engineering from Pohang University of Science and Technology (POSTECH), Korea, in 2007 and 2009, respectively. In 2009, he joined Solid Technology, Inc., and since then he has been engaged in the research and development of the metropolitan pole repeater.

Jong-Ju Hong received his B.S. degree in electronic and electrical engineering from Pohang University of Science Technology (POSTECH), Korea, in 2008. Since 2008, he has been a Research Assistant at the Department of Electronic and Electrical Engineering, POSTECH, where he is currently working toward his M.S. degree. His research interests include multimedia signal processing for display and image signal processing.

Tae Wook Lee is a Senior Research Engineer at LG Display, Timing Controller Development Team, Paju, Korea. He received the B.S., M.S., and Ph.D. degrees from University of Ulsan, Korea in 1998, 2000, and 2004, respectively. His Ph.D. work was on the inner product optimization and its application to image compression. In 2004, he joined LG Display Co. and since then he has worked on the backlight driving, LED local dimming, and timing controller for LCD TVs.

Chang Gone Kim is a Chief Research Engineer at LG Display, Timing Controller Development Team, Paju, Korea. He received his B.S. and M.S. degrees from Kyungbook National University of Korea. Now, he is responsible for designing timing controllers as a Leader at the Timing Controller Development Team, LG Display R&D Center.

Woo-Jin Song received his B.S. and M.S. degrees in electronics engineering from Seoul National University in 1979 and 1981, respectively, and his Ph.D. degree in electrical engineering from Rensselaer Polytechnic Institute in 1986. During 1981–1982, he worked at Electronics and Telecommunication Research Institute (ETRI), Korea. In 1986, he was employed by Polaroid Corp. as a senior engineer, working on digital image processing. In 1989, he was promoted to principal engineering at Polaroid. In 1989, he joined the faculty at Pohang University of Science and Technology (POSTECH), Korea, where he is a professor of electronic and electrical engineering. His current research interests are in the area of digital signal processing; in particular, radar signal processing, signal processing for digital television and multimedia products, and adaptive signal processing.

Journal of the SID 17/12, 2009

1057