054 IEEE Transactions on Consumer Electronics, Vol. 56, o. 4, ovember 0 A Local Dimming Algorithm for Low Power LCD TVs Using Edge-tpe LED Backlight Hunsuk Cho, Student ember, IEEE and Oh-Kong Kwon, ember, IEEE Abstract A local dimming algorithm is proposed to reduce the power consumption of LCD TVs using edge-tpe LED backlight. The proposed local dimming algorithm is implemented b matrix driving of orthogonall located LEDs at the edge of the backlight. The dimming rate of LED string is determined according to target luminance of local blocks and modified to reduce image distortion. The backlight power saving ratios of four tpical images are 3.-53. % and 30.- 54.3% b using proposed division and rate algorithms, respectivel. PSR is greater than 48 db using proposed local dimming algorithms 1. Index Terms Backlight dimming, Local dimming, LCD, Edge-tpe backlight, I. ITRODUCTIO Liquid crstal displa (LCD) TVs has man advantages such as light weight, thin form factor, and high resolution. However, there are some drawbacks that are large power consumption of the backlight and low contrast ratio. Generall LCD displas images b controlling the transparenc of liquid crstal (LC) with backlight of constant luminance. The large backlight power is consumed even in the dark image. The contrast ratio is poor because the light is not blocked completel at the lowest gra level. To solve these problems, backlight dimming algorithms have been studied [1]-[5]. The dimming algorithm is defined as the dnamic control of the luminance of the backlight according to input image. If input image is dark, the power consumption of the backlight can be reduced and contrast ratio can be improved through decrease of the luminance of the backlight. The backlight structure of LCD TVs is divided into direct and edge tpes according to position of light sources as shown in Fig. 1. In direct tpe backlight, light sources are placed on the back of optical films and LC panel as shown in Fig. 1. Because direct-tpe backlight has man numbers of local blocks, the efficienc of local dimming is higher than edge tpe backlight. However, the thickness of backlight is thick because a certain level of distance between light source and optical films is needed for achieving the uniform distribution of the luminance. LC panel Optical films Reflector Fig. 1. Backlight structures of LCD TVs. direct-tpe backlight and edge-tpe backlight. In edge tpe backlight, light sources are placed at the edges of the backlight as shown in Fig. 1. The light is distributed from edges to whole panel through light guide. The thickness of edge-tpe backlight is thinner than that of direct-tpe backlight because of light guide. Like direct-tpe backlight, the local dimming algorithm can be implemented in edge-tpe backlight, [7]. In segment algorithm, LEDs are placed at horizontal edges of the backlight as shown in Fig.. Because one LED string drives one local block like directtpe backlight, the local dimming algorithm can be implemented simpl. However, the efficienc of local dimming is not good because the number of local blocks is small. In Sorted Sector Covering (SSC) algorithm [7], LEDs are placed at four edges of the backlight as shown in Fig.. 1 This work was supported b the IT R&D program of KE/KEIT. [K03051, PIC and BIC for displa and mobile applications] Hunsuk Cho and Oh-Kong Kwon are with Division of Electrical and Computer Engineering, Hanang Universit, 17 Haengdang-dong, Seongdong-gu, Seoul, 133-791, Korea (phone: +8--0-0359; fax: +8-- 97-31; e-mail: okwon@hanang.ac.kr). Contributed Paper anuscript received /15/ Current version published 1/3/ Electronic version published 1/30/. 0098 3063//$0.00 0 IEEE Fig.. The backlight structures of previous local dimming algorithm. segment algorithm and SSC algorithm.
H. Cho and O.-K. Kwon: A Local Dimming Algorithm for Low Power LCD TVs Using Edge-tpe LED Backlight 055 The backlight is divided some sectors according to influence of LED string. The sectors a, b, c, d, e, and f are mainl influenced b one LED string and g, h, i, j, k, l, m, n, and o are influenced b several LED strings. The dimming rate of LED string is determined b comparison between backlight distribution of sectors and luminance of the pixel of input image. However, SSC algorithm has a drawback of complex calculation because backlight distribution of all pixels should be considered. In this paper, we propose a new local dimming algorithm for LCD TVs using edge-tpe backlight. The luminance of local block is determined b sum of the luminance of orthogonall located LEDs at the edge of the backlight. The dimming rate of LED string is calculated for displaing the target luminance of the backlight. To reduce image distortion b wide light distribution of edge-tpe backlight, the dimming ratio of LED string is modified. The effect of the proposed local dimming algorithms is verified b simulation results. II. PROPOSED LOCAL DIIG ALGORITH In this paper, LEDs are placed at four edges of the backlight as shown in Fig. 3. The local dimming algorithm is implemented b matrix driving of orthogonall located LEDs at the edge of the backlight. The luminance of local block is expressed b sum of light emitted from all LEDs. The orthogonall located LEDs provide the local block with 70-85% of light according to the position of the local block and other LEDs provide the remaining light. Fig. 3. Bness of local block b matrix driving of orthogonall located LEDs at the edge of the backlight. The process of proposed local dimming algorithm is shown in Fig. 4. Firstl, the input image stored in memor and characteristic data that are average and maximum luminance of input image are extracted. Secondl, the target luminance of the backlight is determined b using the characteristic data of local block. Thirdl, the dimming rate of LED string is calculated b using the target luminance of the backlight. Fourthl, to reduce image distortion of displaed image, the dimming rate of LED string is modified b partial calculation of image distortion. Fifthl, pixel data of input image are compensated according to backlight distribution. At last, the mixture of backlight and compensated image is displaed on the panel. Fig. 4. The process of the proposed local dimming algorithm A. Calculation of dimming rate of LED string The target luminance of the backlight is determined b using characteristic data of input image. The determination algorithm that is used in direct-tpe backlight can be applied to proposed local dimming sstem. In this paper, correction algorithm [1] is used. In correction algorithm, the target luminance of the backlight is calculated b using the relationship of two parameters that are average and maximum luminance of local block. In the matrix driving, the major part of the luminance of local block is determined b sum of light emitted from orthogonall located LEDs as shown in Fig. 3. In this paper, two calculation algorithms of dimming rate of LED string are proposed. The first calculation method is named b division method. The dimming rates of LED strings at vertical and horizontal edge are expressed as vertical horizontal ax i,1), i,),, i, ), (1) ax ( 1,,,, j) () i, j) is the normalized target luminance of local block (i, j) and and are local block matrix. The second method is named b rate method. In rate method, the dimming rate of LED string is determined b line characteristic of target luminance of the backlight. The dimming rates of LED string at the vertical edge in the side are expressed as ( ax i,1), i,),, i, ) SR, (3)
056 SR 1 k1 sgn Avg row Avg row is the rate coefficient, sqn(z) is sign function that extracts the sign of z, and Avg row is average target luminance of the backlight at row line. The dimming rate of LED string at the vertical edge in side is expressed as ax ( i,1), i,),, i, ) SR, ( ) 1 SR sgn ( ) ( ) 1 i Avgrow i Avgcol k k. The dimming rate of LED string at the horizontal edge in side is expressed as ( j) ax ( 1,,,, j) SR( SR 1 sgn Avgcol Avgrow( k) k1 is the rate coefficient and Avg col is average target luminance of the backlight at column line. The dimming rates of LED string at the horizontal edge in side is expressed as (4) (5) (6) (7) (8) IEEE Transactions on Consumer Electronics, Vol. 56, o. 4, ovember 0 luminance of the backlight is not over the luminance of input image, the displaed image is distorted. To reduce image distortion b wide light distribution of edge-tpe backlight, the modification algorithm of dimming rate of LED string is proposed. In proposed algorithm, the dimming rate of LED string is modified b using image distortion of displaed image. Because the luminance of displaed image cannot exceed the luminance of the backlight [8], the image distortion at pixel ( is expressed as L( LB( (11) L( is normalized maximum luminance of R, G, B sub-pixels of input image at pixel ( and LB( is calculated b the sum of the attenuated light emitted from LEDs to the pixel ( and can be expressed as LB( k1 k1 k1 l k1 r t b (1) l, r, t, and b are coefficients of the attenuation of the light emitted from,,, and LEDs at the edge of the backlight to the pixel (, respectivel. SR ax ( 1,,,, j) SR, ( ) 1 j sgn Avg Avg. col row 1 k (9) () In rate algorithm, if the row line of the backlight is darker than column line of the backlight, unnecessar power consumption can be reduced b decrease of the dimming rate of LED string at vertical edge. B. odification of dimming rate of LED string In edge-tpe backlight, light emitted from LED at the edge of the backlight spread all over the panel. Each of the pixels in local block is influenced not onl orthogonall located LEDs at the edge of the backlight but also other LEDs as shown in Fig. 3. If the dimming rates of neighbor LEDs are small, the luminance of local block can be darker than expectation. If the Fig. 5. odification of dimming rate of LED strings. concerned pixels that have maximum luminance in sub-block and sequence of modification of dimming rate according to average luminance of local blocks In high resolution image, it is hard to calculate the image distortion at all pixels. So the image distortions of specific pixels that greatl influence the image qualit are calculated. The local block is divided into nine sub-blocks for analzing the
H. Cho and O.-K. Kwon: A Local Dimming Algorithm for Low Power LCD TVs Using Edge-tpe LED Backlight 057 influence of neighbor LEDs as shown in Fig. 5. The image distortions of the pixels that have maximum luminance in subblock are calculated. The maximum value of image distortion of concerned pixels in local block is used for modification of dimming rate of LED string. The dimming rates of orthogonall located LED strings are modified according to position of concerned pixel as shown in (13) and (14). If else If else x m / ( j) n / 1 1, 1 1 (13) (14) m and n are m n resolution of input image. At dark local block, the image distortion can be reduced b increasing the dimming rates of orthogonall located LED strings. However, at b local block, the main reason of image distortion is a shortage of light emitted from neighbor LEDs. So, the sequence of modification of dimming rate is determined in order of average luminance of local blocks as shown in Fig. 5. The image distortion at b local block is reduced b suppling light emitted from neighbor LEDs through modification of dimming rate of LED string at dark local block. The backlights b using proposed division and rate algorithms with modification of dimming rate of LED string are shown in Fig. 6 and Fig. 7. 16.9%, 9.5%, and 34.8% b using segment, SSC, division, and rate algorithms, respectivel. The division and rate algorithms save backlight power larger than other algorithms. (c) (d) Fig. 6. The dimmed backlight b using division algorithm. input image, backlight bness b located LEDs at vertical edge, (c) backlight bness b located LEDs at horizontal edge, and (d) backlight bness b all LEDs. III. SIULATIO RESULTS To analze the performance of local dimming algorithm for edge-tpe backlight, four sample images are used as shown in Fig. 8. Two images of flower and residence are used to test influence of luminance, sunrise image is used to test high contrast ratio image, and still-life image is used to test influence of color characteristic. The sample images are full- HD (80 190) resolution and 8 bit gra scale. The edgetpe LED backlight uses 48 LED strings for 8 16 local dimming sstem. The simulated images of segment, CCS, division, and rate algorithms are shown in Fig. 9. Table I shows the backlight power saving ratio of local dimming algorithms. For the low luminance image (flower), the backlight powers are saved 54.9%, 40.4%, 53.%, and 54.3% b using segment, SSC, division, and rate algorithms, respectivel. The algorithms except SSC algorithm save a lot of backlight power more than 50%. For the high luminance image (residence), the backlight powers are saved 11.5%, 1.0%, 3.%, and 30.% b using segment, SSC, division, and rate algorithms, respectivel. The rate algorithm has the largest backlight power saving ratio. For the high contrast ratio image (sunrise), the backlight powers are saved 19.1%, (c) (d) Fig. 7. The dimmed backlight b using rate algorithm. a) input image, backlight bness b located LEDs at vertical edge, (c) backlight bness b located LEDs at horizontal edge, and (d) backlight bness b all LEDs. (c) (d) Fig. 8. Sample images. low luminance image (flower), high luminance image (residence), (c) high contrast ratio image (sunrise), and (d) colorful image (still-life).
058 For the colorful image (still-life), the backlight powers are saved 36.9%, 4.1%, 40.4%, and 45.7% b using segment, SSC, division, and rate algorithms, respectivel. In rate algorithm, the power consumption of the backlight goes down b nearl half. As a result, the rate algorithm saves the largest backlight power and the backlight power saving ratio of SSC algorithm is the lowest. TABLE I COPARISO OF BACKLIGHT POWER SAVIG RATIO Images Low luminance (flower) High luminance (residence) High contrast ratio (sunrise) Colorful (still-life) Images Low luminance (flower) High luminance (residence) High contrast ratio (sunrise) Colorful (still-life) SSC [7] Division Rate 54.9 % 40.4 % 53. % 54.3 % 11.5 % 1.0 % 3. % 30. % 19.1 % 16.9 % 9.5 % 34.8 % 36.9 % 4.1 % 40.4 % 45.7 % TABLE II COPARISO OF PSR OF SIULATED IAGES SSC [7] Division Rate 66 db 66 db 66 db 66 db 53 db 53 db 51 db 50 db 59 db 59 db 55 db 5 db 59 db 61 db 55 db 48 db IEEE Transactions on Consumer Electronics, Vol. 56, o. 4, ovember 0 To compare the image qualit of simulated images, the distribution functions of image distortion and PSR (Peak Signal to oise Ratio) is used. The distribution functions of image distortion are shown in Fig. and Table II shows PSR of simulated images. If PSR is greater than 30 db, it is hard to know the decrease of image qualit [9], []. In segment and SSC algorithms, the values of PSR are 53-66 db. In division algorithm, the values of PSR are 51-66 db. In Rate algorithm, the values of PSR are 48-66 db. As a result, in segment and SSC algorithm, the image qualit is good. However, the backlight power saving ratio is small. The division algorithm saves large backlight power in various images. In rate algorithm, the largest backlight power saving ratio is achieved b keeping image qualit at certain levels. IX. COCLUSIO We have proposed a local dimming algorithm for low power LCD TVs using edge-tpe LED backlight. The high efficient local dimming sstem is implemented using matrix driving of orthogonall located LEDs at the edge of the backlight. The dimming rate of LED string is determined according to target luminance of local blocks and modified to reduce image distortion of displaed image. The power saving ratios of four tpical images are 3.-53. % and 30.-54.3% b using proposed division and rate algorithms, respectivel. SSC [7] Division Rate (c) (d) Fig. 9. Simulated images. low luminance image (flower), high luminance image (residence), (c) high contrast ratio image (sunrise), and (d) colorful image (still-life).
H. Cho and O.-K. Kwon: A Local Dimming Algorithm for Low Power LCD TVs Using Edge-tpe LED Backlight 059 SSC [7] Division 5 3 Image distortion (gra 5 3 Image distortion (gra Rate 5 3 Image distortion (gra 5 3 Image distortion (gra (c) (d) Fig.. Distribution functions of image distortion. low luminance image (flower), high luminance image (residence), (c) high contrast ratio image (sunrise), and (d) colorful image (still-life). PSR is greater than 48 db b using proposed local dimming algorithms. The proposed local dimming algorithms effectivel have higher power saving ratio at the high luminance, high contrast ratio, and colorful images, while PSR of the proposed algorithms is kept similar level with segment and SSC algorithms. REFERECES [1] H. Cho and O. Kwon, A Backlight Dimming Algorithm for Low Power and High Image Qualit LCD Applications, IEEE Trans. Consumer Electron., vol. 55, no., pp. 839-844, a 009. [] T. Funamoto, T. Kobaashi, and T. urao, High-Picture-Qualit Technique for LCD Television: LCD-AI, Proc. International Displa Workshop, pp. 1157-1158, 000. [3]. Raman and G. J. Hekstra, Content Based Contrast Enhancement for Liquid Crstal Displas with Backlight odulation, IEEE Trans. Consumer Electron., vol. 51, no. 1, pp. 18-1, Feb. 005. [4] C. Lai and C. Tsai, Backlight Power Reduction and Image Contrast Enhancement Using Adaptive Dimming for Global Backlight Applications, IEEE Trans. Consumer Electron., vol. 54, no., pp. 669-674, a 008. [5] J. Hong, S. Kim, and W. Song, A Clipping Reduction Algorithm Using Backlight Luminance Compensation for Local Dimming Liquid Crstal Displas, IEEE Trans. Consumer Electron., vol. 56, no. 1, pp. 40-46, Feb. 0. H. Hulze and P. Greef, Power Savings b Local Dimming on a LCD Panel with Side Lit Backlight, SID Int l Smposium Dig. Tech. Papers, pp. 749-75, 009. [7] T. Jung,. Albrecht, C. Xu, and A. Karrenbauer, Application of the SSC Local Dimming Algorithm for an Edge-Lit TV, SID Int l Smposium Dig. Tech. Papers, pp. 1450-1453, 0.
060 [8] L. Kerofsk and S. Dal, Bness Preservation for LCD Backlight Reduction, SID Int l Smposium Dig. Tech. Papers, pp. 14-145, 006. [9] T. Chen, C. Chang, and. Hwang, A Virtual Image Crptosstem Based upon Vector Quantization, IEEE Trans. Image Processing, vol. 7, issue., pp. 1485-1488, Oct. 1998. [] S. Kang and Y. Kim, Image Integrit-based Gra-level Error Control for Low Power Liquid Crstal Displas, IEEE Trans. Consumer Electron., vol. 55, no. 4, pp. 401-406, ov. 009. BIOGRAPHIES Hunsuk Cho received the B.S. degree in electronic engineering from Hanang Universit, Seoul, Korea, in 007. He is currentl pursuing the Ph. D. degree at the same universit. He has been engaged in research for liquid crstal displas. His research interests include image processing and OLED displa. He has authored 3 international journal and conference paper. IEEE Transactions on Consumer Electronics, Vol. 56, o. 4, ovember 0 Oh-Kong Kwon (S 83-88) received the B.S. degree in electronic engineering from the Hanang Universit, Seoul, Korea, in 1978 and the.s. and Ph.D. degrees in electrical engineering from the Stanford Universit in 1986 and 1988, respectivel. From 1980 to 1983, he was with LG Electronics Inc., Seoul, Korea, he involved in the development program of telecommunication products including G-3 fax sstem and PC sstem. From 1987 to 199, he was with the Semiconductor Process and Design Center, Texas Instruments Inc., Dallas, Texas, he was engaged in the development of multi-chip module (C) technologies and smart power integrated circuit technologies for automotive and flat panel displa applications. In 199, he joined Hanang Universit, Seoul, Korea, as an assistant professor in the Department of Electronic Engineering, and is now a distinguished professor in the Division of Electrical and Computer Engineering at Hanang Universit. Dr. Kwon had been served the position of the dean of the College of Engineering at Hanang Universit from 008 to 0. He is now the Provost of the College of Engineering, Hanang Universit. He was an IEEE IED subcommittee member on solid state devices from 1997 to 1998, a technical program chair of 1999 IEEE International Conference on VLSI and CAD, and a workshop co-chair of 000 and 001 Asia-Pacific Workshop on Fundamental and Application of Advanced Semiconductor Devices (AWAD). Dr. Kwon is serving the position of IEEE EDS Korea Chapter Chair. Dr. Kwon was the program manager of Korean TFT-LCD Research and Development Program from 1993 to 1997. Dr. Kwon was the program manager of Korean Flat Panel Research and Development Program from 1998 to 001. Dr. Kwon was a technical program chair of International SoC (Sstem-on-a-Chip) Conference 004, and a technical program chair of International eeting on Information Displas/International Displa anufacturing Conference 006. He is currentl serving in the position of Vice-President of IEEK (Institute of Electronics Engineers of Korea), in the position of an executive chair of International eeting on Information Displa 007, and in the position of a technical program committee member of SID (Societ for Information Displas), ISSCC (International Solid State Circuit Conference) and International Displa Workshop. His research interests include interconnect and electrical noise modeling for high-speed sstem-level integration, waferscale chip-size packages, smart power integrated circuit technologies, mixed mode signal circuit design, and the driving methods and circuits for flat panel displas. He has authored and co-authored over 16 international journal and conference papers and 18 U.S. patents.