低照度环境下的成像，存在亮度低、细节模糊、对比度低且含有大量噪声等问题，会严重影响人眼及机器对图像信息的辨识以及后续对于有用信息的获取。因此，研究低照度图像增强算法有十分重要的意义。本文提出了两种基于Retinex理论的改进算法，具体研究内容为：

⑴针对低照度图像增强算法存在的色彩失真，光晕伪影等问题，本文提出了一种基于图像融合的低照度彩色图像增强方法。该方法在HSI色彩空间对I通道进行处理，然后采用线性色彩恢复算法将图像从HSI转回RGB色彩空间，从而有效的避免了色彩失真。其次，该方法采用加权引导滤波代替了传统Retinex算法中的高斯滤波，由于加权引导滤波各向异性特性，能够更好的对光照分量进行估计。然后，将估计出的光照分量进行复制，对其中一个光照分量采用自适应光照调整函数进行亮度提升；对另一个光照分量采用S型双曲正切函数拓宽灰度动态范围，达到对比度提升的目的；最后利用PCA方法求解权值，并从两幅图像中提取细节特征进行加权融合。实验结果表明，经该方法增强后的图像在局部对比度改善和保持图像的视觉自然性之间进行了平衡，与经典算法相比，该方法可以提高图像的整体亮度和对比度，同时减少低照度环境对成像效果的影响。

⑵现有基于Retinex的图像增强算法在对光照不均匀或者过暗条件下的图像增强后存在边缘模糊，细节纹理不突出以及噪声没有得到有效消除等现象。针对以上问题，本文提出采用梯度域引导滤波和多尺度细节提升算法对Retinex算法进行改进。首先，将输入图像转换至HSI色彩空间，有效的避免了色彩失真。其次，采用梯度域引导滤波估计光照并去噪。梯度域引导滤波是一种具有一阶边缘感知特性的滤波，因此，可以在保持边缘细节的同时有效的避免局部模糊、光晕等问题。最后，引入多尺度细节提升增强图像暗处细节，由于多尺度细节提升算法对图像的三种不同细节层进行加权融合，使得图像暗处细节得到有效提升。实验结果表明，该方法在边缘保持、噪声消除和暗处细节提升方面效果明显。

Low-illumination imaging, with its low brightness, blurred details, low contrast and high noise content, can seriously affect the recognition of image information and subsequent acquisition of useful information by the human eye and machines. Therefore, it is of great importance to study low-light image enhancement algorithms. In this paper, two improved algorithms based on Retinex theory are proposed, with the following details.

(1)In response to the problems of colour distortion and halo artefacts in low- illumination image enhancement algorithms, this paper proposes a low-illumination colour image enhancement method based on image fusion. The method processes the I-channel in the HSI colour space, and then uses a linear colour recovery algorithm to convert the image from HSI back to RGB colour space, thus effectively avoiding colour distortion. Secondly, the method uses a weighted bootstrap filter instead of the Gaussian filter in the traditional Retinex algorithm, which allows for better estimation of the illumination component due to the anisotropic nature of the weighted bootstrap filter. The estimated illumination components are then replicated and an adaptive illumination adjustment function is applied to one of the illumination components for luminance enhancement; an S-type hyperbolic tangent function is applied to the other illumination component to widen the dynamic range of the grey scale for contrast enhancement; finally, the weights are solved using the PCA method and the detailed features are extracted from the two images for weighted fusion. The experimental results show that the images enhanced by this method strike a balance between local contrast improvement and maintaining the visual naturalness of the images, and that the method can improve the overall brightness and contrast of the images compared with classical algorithms, while reducing the impact of low-illumination environments on the imaging effect.

(2)The existing Retinex-based image enhancement algorithm suffers from blurred edges, unremarkable detail texture and ineffective noise elimination after image enhancement under uneven illumination or too dark conditions. To address the above problems, this paper proposes to improve the Retinex algorithm by using gradient domain guided filtering and multi-scale detail enhancement algorithm. Firstly, the input image is converted to HSI color space, which effectively avoids color distortion. Secondly, gradient domain guided filtering is used to estimate the illumination and remove the noise. Gradient domain bootstrap filter is a kind of filter with first-order edge-aware characteristics, so it can effectively avoid local blur and halo while maintaining edge details. Finally, multi-scale detail enhancement is introduced to enhance the image dark details. Since the multi-scale detail enhancement algorithm performs weighted fusion of three different detail layers of the image, the image dark details are effectively enhanced. The experimental results show that the method is effective in edge preservation, noise elimination and dark detail enhancement.

[1]唐永鹤, 蒋烈辉, 侯一凡, 王瑞敏. 基于Retinex与STFT的非接触指纹图像增强[J].计算机工程, 2017,43(02):248-251+256.

[2]李盼池, 孙昊. 量子彩色图像的频域滤波[J]. 电子与信息学报, 2018, 40(03):594-601.

[3]Zong, Xuli, Laine, et al. Speckle Reduction and Contrast Enhancement of Echocardiograms via Multiscale Nonlinear Processing.[J]. IEEE Transactions on Medical Imaging, 1998, 17(4):532-532.

[4]Russo, F. An image enhancement technique combining sharpening and noise reduction[J]. IEEE Transactions on Instrumentation and Measurement, 2001, 51(4):824-828.

[5]于天河, 孟雪, 潘婷, 兰朝凤. 小波变换和自适应变换相结合的图像增强方法[J]. 哈尔滨理工大学学报,2018,23(06):100-104.

[6]宋瑞霞, 李达, 余建德. DT-CWT和色调映射相结合的低照度图像增强算法[J]. 计算机辅助设计与图形学学报,2018,30(07):1305-1312.

[7]海洁. 小波变换图像多聚焦模糊特征自动增强仿真[J].计算机仿真,2019,36(11):360-364.

[8]Zuiderveld K. Contrast Limited Adaptive Histogram Equalization [J].Graphics Gems, 1994: 474-485.

[9]Kim J Y , Kim L S, Hwang S H. An advanced contrast enhancement using partially overlapped sub-block histogram equalization[C]// 2000 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2002.

[10]Ibrahim H, Kong N. Brightness Preserving Dynamic Histogram Equalization for Image Contrast Enhancement.[J]. IEEE Transactions on Consumer Electronics, 2007.

[11]何畏. 基于改进直方图的低照度图像增强算法[J]. 计算机科学,2015,42(S1):241-242+262.

[12]戴声奎, 钟峥, 黄正暐. 基于最大熵模型的双直方图均衡算法[J].电子学报,2019,47(03):678-685.

[13]曹海杰,刘宁,许吉,彭杰,刘宇昕.红外图像自适应逆直方图增强技术[J].红外与激光工程,2020,49(04):256-262.

[14]Ancuti C O, Ancuti C. Single Image Dehazing by Multi-Scale Fusion[J]. IEEE Transactions on Image Processing, 2013, 22(8):3271-3282.

[15]Ancuti C, Ancuti C O, Haber T, et al. Enhancing underwater images and videos by fusion[C]// IEEE Conference on Computer Vision & Pattern Recognition. IEEE, 2012.

[16]Ancuti C O, Ancuti C, Vleeschouwer C D, et al. Color Balance and Fusion for Underwater Image Enhancement[J]. IEEE Transactions on Image Processing, 2017, 27(99):379-393.

[17]Li Z, Wei Z, Wen C, et al. Detail-Enhanced Multi-Scale Exposure Fusion[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2017, 26(3):1243-1252.

[18]Ying Z, Ge L, Wen G. A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement[J]. 2017.

[19]Zhang Q, Nie Y, Zheng W S. Dual Illumination Estimation for Robust Exposure Correction[J]. John Wiley & Sons, Ltd, 2019.

[20]Wang W, Chen Z, Yuan X, et al. Adaptive Image Enhancement Method for Correcting Low-Illumination Images[J]. Information Sciences, 2019, 496.

[21]Zhou Z, Li S, Wang B. Multi-scale weighted gradient-based fusion for multi-focus images[J]. Information Fusion, 2014, 20:60-72.

[22]Xueyang, Paisley, John, et al. A fusion-based enhancing method for weakly illuminated images[J]. Signal Processing: The Official Publication of the European Association for Signal Processing (EURASIP), 2016.

[23]Ma K, Duanmu Z, Yeganeh H, et al. Multi-Exposure Image Fusion by Optimizing A Structural Similarity Index[J]. IEEE Transactions on Computational Imaging, 2018, 4(1):60-72.

[24]Ma K, Hui L, Yong H, et al. Robust Multi-Exposure Image Fusion: A Structural Patch Decomposition Approach[J]. IEEE Transactions on Image Processing, 2017, 26(5):2519-2532.

[25]Yu L, Xun C, Ward R K, et al. Image Fusion With Convolutional Sparse Representation[J]. IEEE Signal Processing Letters, 2016, PP(99):1-1

[26]Land E. Lightness and Retinex Theory[J]. Journal of the Optical Society of America, 1971, 61(1):1-11.

[27]Jobson D J, Rahman Z U, Woodell G A . Properties and Performance of a Center/Surround Retinex[J]. IEEE Transactions on Image Processing, 1997, 6(3):451-462.

[28]Petro A B, Sbert C, Jean M M. Multi-scale Retinex [J]. Image Processing on Line, 2014, 5(4): 71-88.

[29]赵宏宇, 肖创柏, 禹晶, 等.马尔科夫随机场模型下的 Retinex夜间彩色图像增强[J].光学精密工程,2014,22(04):1048-1055.

[30]李红, 王瑞尧, 耿则勋, 等. 基于多尺度梯度域引导滤波的低照度图像增强算法[J]. 计算机应用, 2019, v.39;No.350(10):258-264.

[31]田会娟, 蔡敏鹏, 关涛, 等. 基于YCbCr颜色空间的Retinex低照度图像增强方法研究[J]. 光子学报, 2020, v.49(02):173-184.

[32]Shen L, Yue Z, Feng F, et al. MSR-net:Low-light Image Enhancement Using Deep Convolutional Network[J]. 2017.

[33]Wei C, Wang W, Yang W, et al. Deep Retinex Decomposition for Low-Light Enhancement[J]. 2018.

[34]Cai J, Gu S, Zhang L. Learning a Deep Single Image Contrast Enhancer from Multi-Exposure Images[J]. IEEE Transactions on Image Processing, 2018, 27(4):2049-2062.

[35]Lv F, Lu F, Wu J H, et al. MBLLEN: Low-Light Image/Video Enhancement Using CNNs[C]. 2018 British Machine Vision Conference (BMVC). Newcastle, England, 2018: 1-13.

[36]程宇, 邓德祥, 颜佳, 等. 基于卷积神经网络的弱光照图像增强算法 [J]. 计算机应用, 2019, 39(04): 1162-1169.

[37]马红强, 马时平, 许悦雷, 等. 基于深度卷积神经网络的低照度图像增强[J]. 光学学报, 2019,39(02): 99-108.

[38]Fu X, Zeng D, Yue H, et al. A fusion-based enhancing method for weakly illuminated images[J]. Signal Processing, 2016, 129(dec.):82-96.

[39]Frankle J A, Mccann J J. Method and apparatus for lightness imaging[J]. US, 1983.

[40]Jobson D J, Rahman Z, Woodell G A. A multiscale retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing, 2002, 6(7):965-976.

[41]Guo X, Yu L, Ling H. LIME: Low-light Image Enhancement via Illumination Map Estimation[J]. IEEE Transactions on Image Processing, 2016, PP(99):1-1.

[42]Cai B, Xu X, Guo K, et al. A Joint Intrinsic-Extrinsic Prior Model for Retinex[C]// 2017 IEEE International Conference on Computer Vision (ICCV). IEEE, 2017.

[43]Shuhang, Wang, Jin, et al. Naturalness preserved enhancement algorithm for non-uniform illumination images.[J]. IEEE transactions on image processing : a publication of the IEEE Signal Processing Society, 2013, 22(9):3538-48.

[44]Tomasi C , Manduchi R . Bilateral filtering for gray and color images[C]// International Conference on Computer Vision. IEEE, 2002.

[45]He K,Sun J, Tang X. Guided Image Filtering[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2013, 35(6):1397-1409.

[46]Li Z, Zheng J, Zhu Z, et al. Weighted Guided Image Filtering[J]. IEEE Transactions on Image Processing, 2014, 24(1).

[47]黄黎红. 一种基于单尺度Retinex的雾天降质图像增强新算法[J]. 应用光学, 2010, 31(05): 728-733.

[48]赵全友, 潘保昌, 郑胜林, 等. 一种颜色保持的彩色图像增强新算法[J]. 计算机应用, 2008(02): 448-451.

[49]Lee S, Han H, Kwak B, et al. Color image enhancement method using a space-variant luminance map[C]// Digest of Technical Papers International Conference on Consumer Electronics. IEEE, 2010.

[50]Dai H , Gu X , Lei H, et al. An adaptive color image enhancement algorithm based on local luminance discontinuity. IEEE, 2010.

[51]Gorai A, Ghosh A . Hue-preserving color image enhancement using particle swarm optimization[M]. IEEE, 2011.

[52]Kou F, Chen W, Wen C, et al. Gradient Domain Guided Image Filtering[J]. IEEE Transactions on Image Processing, 2015, 24(11):4528-4539.

[53]Sun X, Liu H, Wu S, et al. Low-Light Image Enhancement Based on Guided Image Filtering in Gradient Domain[J]. International Journal of Digital Multimedia Broadcasting, 2017, 2017:1-13.

[54]涂清华, 戴声奎. 基于域滤波的自适应Retinex图像增强[J]. 计算机工程与科学, 2016, 38(09): 1830-1835.

[55]Kim Y, Koh Y J, Lee C, et al. Dark image enhancement based onpairwise target contrast and multi-scale detail boosting[C]// IEEE International Conference on Image Processing. IEEE, 2015:1404-1408.

[56]D Xuan,Pang Y A,Wen J G. Fast efficient algorithm for enhancement of low lighting video[C]// IEEE International Conference on Multimedia & Expo. ACM, 2011.

[57]Y. Zhang, J. Zhang, and X. Guo, “Kindling the darkness: A practical low-light image enhancer,” in Proceedings of the 27th ACM International Conference on Multimedia, 2019, pp. 1632–1640.