为均衡增强低照度图像的同时,保留其更多的细节信息,提出一种改进Retinex低照度图像增强算法. 该算法基于HSV(Hue, Saturation, Value)颜色空间,对分离出的明度分量和饱和度分量进行增强. 首先,使用限制对比度自适应直方图均衡化(Contrast Limited Adaptive Histogram Equalization,CLAHE)优化明度分量,使图像更接近均匀光照场景,并使用自适应Gamma对饱和度分量进行校正. 然后,采用三维块匹配滤波(Block-matching and 3D Filtering, BM3D)算法对光照分量进行估计,并求得相应的反射分量,提出一种改进Gamma变换函数,依据光照分量信息对明度分量进行增强,同时,采用Gabor滤波器和Canny算法对原图进行细节提取,提出一种细节增强策略,对反射分量及其纹理细节进行增强. 最后,将各分量进行加权融合,再将增强图像变换回RGB空间. 实验结果表明,所提算法相较于自动色彩均衡、自适应局部色调映射、低光照图像增强、带色彩恢复多尺度视网膜增强算法有更好的增强效果和普适性,且原图经过增强后,信息熵、峰值信噪比、结构相似性指数、图像质量指数、平均梯度有显著提升,均方根误差显著下降.
Abstract
An improved Retinex low-illumination image enhancement algorithm is proposed for the balanced enhancement of low-illumination images while retaining their more detailed information. The algorithm is based on the HSV (Hue, Saturation, Value) color space and enhances the separated luminance and saturation components. First, the brightness component is optimized using Contrast Limited Adaptive Histogram Equalization (CLAHE) to make the image closer to the uniformly illuminated scene, and the saturation component is corrected using Adaptive Gamma. Then, a Block-matching and 3D Filtering (BM3D) algorithm is used to estimate the illumination component, the corresponding reflection component is obtained, and an improved Gamma transform function is proposed to enhance the luminance component based on the information of the illumination component. Meanwhile, the Gabor filter and Canny algorithm are used to extract the details of the original image, and a detail enhancement strategy is proposed to enhance the reflection component and its texture details. Finally, the components are fused with multiple weights, and then the enhanced image is transformed back to RGB space. Experimental results show that the proposed algorithm has better enhancement effect and universality than automatic color equalization, adaptive local tone mapping, low-illumination image enhancement, and multi-scale Retinex with color restoration. After enhancement, the original image showed significant improvements in information entropy,peak signal-to noise ratio, structural similarity index,universal image quality index,average gradient,while the root mean square error decreased significantly.
从数据集中随机抽取200张多场景低照度图像作为实验数据,包含只存在暗部信息的图像100张,光、暗信息共存的图像100张,图像场景包含室内、室外、人物、自然及城市景观等,可从不同角度验证改进方法的有效性,并使用自动色彩均衡(Automatic Color Enhancement,ACE)、自适应局部色调映射(Adaptive Local Tone Mapping,ALTM)、低光照图像增强(Low-light Image Enhancement,LIME)、带色彩恢复多尺度视网膜增强(Multi-scale Retinex with Color Restoration,MSRCR)算法同本文方法BG-Retinex对抽取的低照度图像进行增强,实验在Pycharm 2021、Windows10、CPU Intel Core i9-10900、RAM 32 G测试平台上进行. 4组实验场景图像如图9所示.
OU B, JIANGX Y, XIONGJ Q .Algorithm of medical image reversible data hiding for contrast enhancement[J].Journal of Hunan University (Natural Sciences),2022,49(4):26-34.(in Chinese)
HANY C, ZHANGW W, HEW J, et al. Low-light true color image enhancement algorithm based on adaptive truncation simulation exposure and unsupervised fusion[J]. Acta Photonica Sinica, 2023,52(9): 0910002. (in Chinese)
[6]
WANGT, YANGJ, JIZ X,et al .Probabilistic diffusion for interactive image segmentation[J].IEEE Transactions on Image Processing,2019,28(1):330-342.
ZHOUL H, GONGJ K, LIB .Image information restoration of automotive strip steel surface based on sparse representation[J].Journal of Hunan University (Natural Sciences),2021,48(8):141-148.(in Chinese)
[9]
WANGK, HUANGF Z .An improved MSRCR low illumination image enhancement algorithm combined with residual fusion[C]//2021 40th Chinese Control Conference (CCC), September 6,2021.Shanghai,China:IEEE,2021:2993-2998.
[10]
XIEY M, NINGL C, WANGM Q,et al .Image enhancement based on histogram equalization[J]. Journal of Physics:Conference Series,2019, 1314(1): 012161.
SUB, LIC, WANGL .Mine image enhancement algorithm based on Retinex using multi-weight fusion strategy[J].Journal of China Coal Society,2023,48(Sup.2):813-822.(in Chinese)
[13]
MUQ, WANGX Y, WEIY Y,et al .Low and non-uniform illumination color image enhancement using weighted guided image filtering[J].Computational Visual Media,2021,7(4):529-546.
YANZ, JIANGL, YANGF,et al .Bi-histogram equalization algorithm for infrared image enhancement[J]. Infrared Technology,2022,44(9):944-950.(in Chinese)
[16]
TIRUMANIV H L, TENNETIM, SRIKAVYA KCH,et al .Image resolution and contrast enhancement with optimal brightness compensation using wavelet transforms and particle swarm optimization[J].IET Image Processing,2021,15(12):2833-2840.
[17]
OHJ G, HONGM C. Adaptive image rendering using a nonlinear mapping-function-based Retinex model[J]. Sensors, 2019, 19(4):969.
LIZ L, WANGH W, CAOW Y,et al .A method for enhancing low light images in coal mines based on Retinex model containing noise[J].Journal of Mine Automation,2023,49(4):70-77.(in Chinese)
WEIH L, WANGJ X .Simulation research on edge sharpening enhancement of motion blurred digital image[J]. Computer Simulation, 2020, 37(7): 459-462.(in Chinese)