基于全变分Retinex及梯度域的雾天图像增强算法

doi:10.3969/j.issn.1000-436x.2014.06.018

摘要/Abstract

摘要：

为提高雾天图像增强的对比度并保持颜色恒常性，提出了基于全变分 Retinex 及梯度域的雾天图像增强算法。首先，采用高斯—赛德尔 GS（Gauss-Seidel）迭代算法对基于 Retinex 的全变分能量泛函数进行求解，从而有效地保持颜色恒常性；其次，采用相对梯度与绝对梯度相结合的方式拉伸雾天图像较亮处的梯度，在全变分Retinex理论下重建增强后的雾天图像，并将该增强算法应用到彩色图像；最后，加权融合基于全变分Retinex增强算法与梯度域增强算法的增强结果，使得增强结果既能提高对比度又能保持色彩恒常性。实验结果表明，本算法提高了雾天图像增强后的对比度和清晰度，具有颜色恒常性、颜色保真高等特性。

关键词: 雾天图像, 图像增强, Retinex算法, 梯度域

Abstract:

Considering the preservation of color constancy and the improvement of contrast ratio of foggy image en-hancement, the foggy image enhancement algorithm based on the total variational Retinex gradient domain method was proposed. Firstly, the algorithm used the GS（Gauss-Seidel）iterative method to resolve the Retinex variational energy functional so that the color constancy was remained and the operational velocity of the proposed algorithm was improved. Secondly, the algorithm stretched strongly the gradient of the higher light blocks in the foggy image by combining the relative gradient and the absolute gradient, and reconstructed the foggy enhancement images with the context of least squares and extended the enhancement method to process the color images. Lastly, the enhancement images by using the total variational Retinex method and the gradient domain algorithm were fused weightedly to keep the color constancy and improve the contrast ratio. The experimental results show that the proposed algorithm can effectively improve the image's contrast and definition,the algorithm could keep better color constancy and high color fidelity.

Key words: foggy image, images enhancement, Retinex algorithm, gradient domain

陈炳权,刘宏立. 基于全变分Retinex及梯度域的雾天图像增强算法[J]. 通信学报, 2014, 35(6): 139-147.

Bing-quan CHEN,Hong-li LIU. Algorithm for foggy image enhancement based on the total variational Retinex and gradient domain[J]. Journal on Communications, 2014, 35(6): 139-147.

图/表 9

图1

图2

图3

图4

图5

图6

图7

表1

表2

参考文献 25

[1]	WILLIAM K P . Digital Image Processimg[M]. Beijing: China Machine Press, 2005.158-196.
[2]	LAND E H . Recent advances in retinex theory and some implications for cortical computations: color vision and the natural image[J]. Pro-ceedings of the National Academy of Sciences of the United States of America, 1983,80 (16): 5163-5169.
[3]	禹晶，徐东彬，廖庆敏 . 图像去雾技术研究进展[J]. 中国图象图形学报， 2011,16(9): 1561-1576. YU J , XU D B , LIAO Q M . Image defogging:a survey[J]. Journal of Image and Graphics, 2011,16(9): 1561-1576.
[4]	WANG H , HE X H , YANG X M . An adaptive foggy image enhance-ment algorithm based on fuzzy theory and clahe[J]. Microelectronics and Computer, 2012,29(1): 32-34.
[5]	WANG R G , FU J F , YANG Z X . A novel retinex algorithm based on dark channel prior mode[J]. Acta Electronica Sinica, 2013,41(6): 1188-1192.
[6]	KO N , LOUIS K , STEPHEN L . Bayesian defogging[J]. Int J Comput Vis, 2012,98:263-278.
[7]	YOON I , KIM S Y , KIM D Y . Adaptive defogging with color correc-tion in the HSV color space for consumer video surveillance sys-tems[A]. IEEE International Conference on Consumer Electronics (ICCE)[C]. 2012.606-607.
[8]	WANG R , YANG X G . A fast method of foggy image enhancement[A]. International Conference on Measurement, Information and Control (MIC)[C]. 2012.883-887.
[9]	JANG J H . Contrast-enhanced fusion of multisensor images using subband-decomposed multiscale retinex[J]. IEEE Transactions on Im-age Processing, 2012,21(8): 3479-3490.
[10]	JIN X B , BAO J , DU J J . Image enhancement based on selective-retinex fusion algorithm[J]. Journal of Software, 2012,7(6): 1187-1194.
[11]	FATTAL R , LISCHINSKI D , WERMAN M . Gradient domain high dynamic range compression[A]. Computer Graphics Proceed-ings,Annual Conference Series, ACM SIGGRAPH, SanAntonio[C]. Texas, 2002.249-256.
[12]	KIMMEL R , ELAD M , SHAKED D , et al. A variational framework for retinex[J]. International Journal of Computer Vision, 2003,52(1): 7-23.
[13]	朱立新，王平安，夏德深 . 基于梯度场均衡化的图像对比度增强[J]. 计算机辅助设计与图形学学报， 2007,19(12): 1546-1552. ZHU L X , WANG P A , XIA D S . Image contrast enhancement by gra-dient field equalization[J]. Journal of Computer-Aided Design＆Computer Graphics, 2007,19(12): 1546-1552.
[14]	李明，杨艳屏 . TV-Retinex：一种快速图像增强算法[J]. 计算机辅助设计与图形学学报， 2010,22(10): 1777-1782. LI M , YANG Y P . TV-Retinex: a fast algorithm for image enhance-ment[J]. Journal of Computer-Aided Design and Computer Graph-ics, 2010,22(10): 1777-1782.
[15]	JOBSON D J , RAHMAN Z , WOODELL G A . Properties and per-formance of a center/Surround retinex[J]. IEEE Transactions on Image Processing, 1997,6(3): 451-462.
[16]	RAHMAN Z , WOODELL G A , JOBSON P J . A comparison of the multiscale retinex with other image enhancement techniques[A]. IS＆T 50th Annual Conference,Final Program and Proceedings[C]. 1997.426-431.
[17]	RUDIN L , OSHER S , FATEMI E . Nonlinear total variation based noise removal algorithms[J]. Phy D, 1992,60:259-268.
[18]	MAJUMDER A , IRANI S . Perception based contrast enhancement of images[J]. ACM Transactions on Applied Perception, 2007,4(3): 17.
[19]	LI M . fast algorithm for color image enhancement with total varia-tion regularization[J]. Science China (Information Sciences), 2010(9): 1913-1916.
[20]	张新龙，汪荣贵，张璇等. 雾天图像增强计算模型及算法[J]. 中国Z图象图形学报， 2011,16(8): 1359-1368. HANG X L , WANG R G , ZHANG X , et al. Calculation model and algorithm in foggy image enhancement[J]. Journal of Image and Graphics, 2011,16(8): 1359-1368.
[21]	汪荣贵，张新龙，张璇等. 基于 Retinex 理论的压缩域图像增强方法研究[J]. 计算机研究与发展， 2011,48(2): 259-270. WANG R G , ZHANG X L , ZHANG X , et al. Image enhancement in the compressed domain based on Retinex theory[J]. Journal of Com-puter Research and Development, 2011,48(2): 259-270.
[22]	LU S L , ZOU L J , SHEN X H , et al. Multi-spectral remote sensing image enhancement method based on PCA and IHS transformations[J]. Journal of Zhejiang University-Science A(Applied Physics＆Engi-neering), 2011(6):453-460.
[23]	许欣，陈强，孙怀江等. 结合视觉感知特性的梯度域图像增强方法[J]. 计算机辅助设计与图形学学报， 2009,21(1): 130-135. XU X , CHEN Q , SUN H J , et al. Perception based Gradient domain enhancement of images[J]. Journal of Computer-Aided Design and Computer Graphics, 2009,21(1): 130-135.
[24]	JAE H J , BOORYM C , SUNG D K . Sub-band de-composed multis-cale retinex with space varying gain[A]. IEEE International Conference on Image Processing[C]. 2008.3168-3171.
[25]	CHAMBAH M , RIZZI A , GATTA C , et al. Perceptual approach for unsupervised digital color restoration of cinematographic ar-chives[A]. Color Image VIII: Processing, Hardcopy, and Applica-tions, CA, USA, 2003. 138-149.

图序	HE算法		全变分Retinex方法		梯度域方法		本文算法
图序	相关系数	巴氏距离	相关系数	巴氏距离	相关系数	巴氏距离	相关系数	巴氏距离
图1	0.009 1	0.461 5	0.252 0	0.301 0	0.457 8	0.281 0	0.851 2	0.121 3
图2	0.007 8	0.402 5	0.235 1	0.321 1	0.522 0	0.247 3	0.896 5	0.102 9
图3	0.005 3	0.312 5	0.135 2	0.215 2	0.231 0	0.201 1	0.452 6	0.113 6

图序	亮度均值	对比度均值	信息熵	清晰度
图2(a)	176.15	12.03	1.809 8	17.17
图2(b)	131.26	38.17	1.718 2	73.83
图2(c)	127.98	12.29	2.100 8	35.10
图2(d)	128.12	15.17	1.754 7	32.71
图2(e)	126.96	26.53	2.161 5	36.56
图3(a)	165.26	11.29	1.500 3	15.22
图3(b)	133.30	33.10	1.625 8	62.25
图3(c)	137.23	12.11	1.743 0	34.37
图3(d)	135.15	13.20	2.002 5	32.69
图3(e)	123.39	27.18	2.268 8	39.84
图4(a)	133.15	10.36	1.611 8	13.17
图4(b)	125.18	35.16	1.720 1	58.16
图4(c)	129.79	13.55	1.613 5	35.16
图4(d)	130.23	12.78	1.918 9	33.59
图4(e)	126.11	19.29	2.179 1	38.10