低照度图像增强是一种通过增强图像整体或局部区域对比度来提高图像质量的重要技术，它能有效改善图像的视觉效果。然而，目前低照度图像增强技术在进行图像增强时仍存在三个问题：一是图像增强技术在增强复杂光照条件下的非均匀照度图像时，会出现过度增强和增强不足的问题，二是低照度图像的暗区域中存在噪声，图像增强技术在提高对比度的同时会放大噪声，三是图像增强技术在处理低信噪比图像时不能较好的处理同为高频的细节信息和噪声。因此，针对上述三个问题，提出了三种有效的低照度图像增强算法，主要研究内容及成果如下：

（1）针对过度增强和增强不足的问题，提出一种基于信息熵的自适应伽马矫正算法。该算法利用伽马函数对图像直方图进行修改使其分布更均匀，以提高图像对比度。算法进一步将图像信息熵作为衡量直方图均匀程度的指标，求解其最大时对应的最优伽马值，并应用于伽马矫正，从而实现算法的自适应能力。实验结果表明，所提算法与其他六种经典的图像增强算法相比，在主观视觉和客观指标两个方面都有较好的效果。

（2）针对图像中噪声被放大的问题，提出一种基于鲁棒性主成分分析（Robust Principal Component Analysis，RPCA）的自适应增强算法。该算法通过RPCA分解将照度信息与噪声分离得到低秩分量和稀疏分量，并对低秩分量采用上述自适应伽马矫正算法提高图像对比度。所提算法充分考虑像素之间的空间关系，且对幅值较大的噪声有较强的鲁棒性。实验结果表明，所提算法与其他四种经典的图像增强算法相比，在主观视觉和客观指标两个方面都有较好的效果。

（3）针对低信噪比图像中存在大量噪声的问题，提出了一种基于降噪自编码网络的自适应增强算法。该算法以降噪自编码器为网络模型，选择RPCA分解模型中的稀疏分量作为训练集，将稀疏分量中同为高频的细节信息和噪声进行分离，实现噪声抑制，并通过分步训练进一步改善网络的噪声抑制效果。实验结果表明，所提算法与其他两种经典的降噪算法相比，在主观视觉和客观指标两个方面都有较好的效果。

Low illumination image enhancement is an important technology to improve image quality by enhancing the overall or partial area contrast of the whole or local area of the image, and it effectively improves the visual effect of the image. However, current low-illuminance images still have three main shortcomings that need to be solved: First, the complex lighting conditions make the image illuminance uneven, and the enhancement results will have over-enhancement or under-enhancement problems. Second, the interference information in the image will increase with the contrast. The increase of is magnified, and thirdly, the noise with larger amplitude will obscure the details of the image. In response to the above three problems, three low-light image enhancement algorithms are proposed. The main research contents and results are as follows:

(1) In order to solve the problem of complex lighting conditions in the image, an adaptive gamma correction algorithm based on information entropy was proposed. The algorithm calculates the optimal parameters of gamma correction by maximizing the information entropy to realize the adaptive adjustment of image contrast, effectively avoiding the problems of over-enhancement and under-enhancement. In addition, the algorithm has higher operating efficiency and better real-time performance. According to experimental comparison and analysis, result show that compared with other six classic image enhancement algorithms, the proposed algorithm has better results in both subjective vision and objective indicators.

(2) Aiming at the problem that the interference information in the image is amplified, an adaptive enhancement algorithm based on robust principal component analysis (RPCA) was proposed. The algorithm uses RPCA decomposition to separate the illuminance information and noise to obtain low-rank components and sparse components, and uses the above-mentioned adaptive gamma correction algorithm for low-rank components to improve image contrast. The proposed algorithm adequately considers the spatial relationship between pixels, and has strong robustness to noise with larger amplitude. According to experimental comparison and analysis, results show that, compared with the other four classic image enhancement algorithms, the proposed algorithm has better results in both subjective vision and objective indicators.

(3) Aiming at the problem of a large amount of noise in low signal-to-noise ratio images, an adaptive enhancement algorithm based on denoising self-encoding network was proposed. The algorithm uses the denoising autoencoder as the network model, and selects the sparse components in the RPCA decomposition model as the training set to separates the detailed information and noise which belong to high-frequency components to realizes noise suppression, and further trains them through step-by-step training. Improve the noise suppression effect of the network. According to experimental comparison and analysis, results show that the proposed algorithm has better results in both subjective vision and objective indicators compared with the other two classic noise reduction algorithms.

[1]Singh N, Kaur L, Singh K. Histogram equalization techniques for enhancement of low radiance retinal images for early detection of diabetic retinopathy[J]. Engineering Science and Technology, an International Journal, 2019, 22(3): 736-745.

[2]Soumya T, Thampi S M. Recolorizing dark regions to enhance night surveillance video[J]. Multimedia Tools and Applications, 2017, 76(22): 24477-24493.

[3]Du Y, Tong M, Zhou L, et al. Edge detection based on Retinex theory and wavelet multiscale product for mine images[J]. Applied Optics, 2016, 55(34): 9625-9637.

[4]Wen S, Hu X, Ma J, et al. Autonomous robot navigation using Retinex algorithm for multiscale image adaptability in low-light environment[J]. Intelligent Service Robotics, 2019, 12(4): 359-369.

[5]Jin Z H U, Li L I, Wei-qi J I N, et al. Natural-appearance Colorization and Enhancement for the Low-light-level Night Vision Imaging[J]. Acta Photonica Sinica, 2018, 47(4): 410002.

[6]李金洪, 邹梅. 低照度CMOS图像传感器设计与实现[J]. 红外与激光工程, 2018, 47(07): 256-262.

[7]张元涛, 柴孟阳, 孙德新, 刘银年. 全局快门sCMOS图像传感器数字TDI微光成像技术[J]. 光学学报, 2018, 38(09): 144-151.

[8]Zhou Z, Sang N, Hu X. Global brightness and local contrast adaptive enhancement for low illumination color image[J]. Optik, 2014, 125(6): 1795-1799.

[9]Xu Q, Jiang H, Scopigno R, et al. A novel approach for enhancing very dark image sequences[J]. Signal processing, 2014, 103: 309-330.

[10]Mu D, Xu C, Ge H. Hybrid genetic algorithm based image enhancement technology[C]//2011 International Conference on Internet Technology and Applications. IEEE, 2011: 1-4.

[11]Srinivas K, Bhandari A K. Low light image enhancement with adaptive sigmoid transfer function[J]. IET Image Processing, 2019, 14(4): 668-678.

[12]李乐鹏, 孙水发, 夏冲, 等. 直方图均衡技术综述[J]. 计算机系统应用, 2014, 23(03): 1-8.

[13]冈萨雷斯[美].数字图像处理(第三版)[M].电子工业出版社，2011.

[14]田小平, 徐小京, 吴成茂. 基于LIP模型的低照度彩色图像增强新算法[J]. 西安邮电大学学报, 2015, 20(01): 9-13.

[15]Li L, Sun S, Xia C, et al. Survey of histogram equalization technology[J]. Comput. Syst. Appl., 2014, 23(3): 1-8.

[16]Land E H, McCann J J. Lightness and retinex theory[J]. Josa, 1971, 61(1): 1-11.

[17]Park S, Kim K, Yu S, et al. Contrast enhancement for low-light image enhancement: A survey[J]. IEIE Transactions on Smart Processing & Computing, 2018, 7(1): 36-48.

[18]Łoza A, Bull D R, Hill P R, et al. Automatic contrast enhancement of low-light images based on local statistics of wavelet coefficients[J]. Digital Signal Processing, 2013, 23(6): 1856-1866.

[19]Sun T, Jung C, Ke P, et al. Readability enhancement of low light videos based on discrete wavelet transform[C]//2017 IEEE International Symposium on Multimedia (ISM). IEEE, 2017: 342-345.

[20]Wang L, Fu G, Jiang Z, et al. Low-light image enhancement with attention and multi-level feature fusion[C]//2019 IEEE International Conference on Multimedia & Expo Workshops (ICMEW). IEEE, 2019: 276-281.

[21]Zhu J, Jin W, Li L, et al. Fusion of the low-light-level visible and infrared images for night-vision context enhancement[J]. Chinese Optics Letters, 2018, 16(1): 013501.

[22]Zhu J, Wang Z. Low-illumination surveillance image enhancement based on similar scenes[J]. Comput. Appl. Softw., 2015, 32(1): 203-205.

[23]张红英, 朱恩弘, 吴亚东. 一种基于细节层分离的单曝光HDR图像生成算法[J]. 自动化学报, 2019, 45(11): 2159-2170.

[24]Yamakawa M, Sugita Y. Image enhancement using Retinex and image fusion techniques[J]. Electronics and Communications in Japan, 2018, 101(8): 52-63.

[25]李赓飞, 李桂菊, 韩广良,等. 亮通道先验Retinex对低照度图像的光照补偿[J]. 光学精密工程, 2018, 26(05): 1191-1200.

[26]Zhang L, Shen P, Peng X, et al. Simultaneous enhancement and noise reduction of a single low-light image[J]. IET Image Processing, 2016, 10(11): 840-847.

[27]Yan Z, Zhang H, Wang B, et al. Automatic photo adjustment using deep neural networks[J]. ACM Transactions on Graphics (TOG), 2016, 35(2): 1-15.

[28]Lore K G, Akintayo A, Sarkar S. LLNet: A deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition, 2017, 61: 650-662.

[29]Park S, Yu S, Kim M, et al. Dual autoencoder network for retinex-based low-light image enhancement[J]. IEEE Access, 2018, 6: 22084-22093.

[30]Loh Y P, Chan C S. Getting to know low-light images with the exclusively dark dataset[J]. Computer Vision and Image Understanding, 2019, 178: 30-42.

[31]Russakovsky O, Deng J, Su H, et al. Imagenet large scale visual recognition challenge[J]. International journal of computer vision, 2015, 115(3): 211-252.

[32]Chen C, Chen Q, Xu J, et al. Learning to see in the dark[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 3291-3300.

[33]Lin T Y, Maire M, Belongie S, et al. Microsoft coco: Common objects in context[C]//European conference on computer vision. Springer, Cham, 2014: 740-755.

[34]Wright J, Ganesh A, Rao S, et al. Robust principal component analysis: Exact recovery of corrupted low-rank matrices via convex optimization[J]. Advances in neural information processing systems, 2009, 22: 2080-2088.

[35]Huang S C, Yeh C H. Image contrast enhancement for preserving mean brightness without losing image features[J]. Engineering Applications of Artificial Intelligence, 2013, 26(5-6): 1487-1492.

[36]Tan T L, Sim K S, Tso C P. Image enhancement using background brightness preserving histogram equalisation[J]. Electronics letters, 2012, 48(3): 155-157.

[37]Singh K, Vishwakarma D K, Walia G S, et al. Contrast enhancement via texture region based histogram equalization[J]. Journal of modern optics, 2016, 63(15): 1444-1450.

[38]TIAN X, QIAO D, WU C. Color image enhancement based on bi-histogram equalization[J]. Journal of Xi'an University of Posts and Telecommunications, 2015.

[39]Parihar A S, Verma O P. Contrast enhancement using entropy-based dynamic sub-histogram equalisation[J]. IET Image Processing, 2016, 10(11): 799-808.

[40]Celik T, Tjahjadi T. Contextual and variational contrast enhancement[J]. IEEE Transactions on Image Processing, 2011, 20(12): 3431-3441.

[41]Lee C, Lee C, Kim C S. Contrast enhancement based on layered difference representation of 2D histograms[J]. IEEE transactions on image processing, 2013, 22(12): 5372-5384.

[42].Huang S C, Cheng F C, Chiu Y S. Efficient contrast enhancement using adaptive gamma correction with weighting distribution[J]. IEEE transactions on image processing, 2012, 22(3): 1032-1041.

[43]Veluchamy M, Subramani B. Image contrast and color enhancement using adaptive gamma correction and histogram equalization[J]. Optik, 2019, 183: 329-337.

[44]Cao G, Huang L, Tian H, et al. Contrast enhancement of brightness-distorted images by improved adaptive gamma correction[J]. Computers & Electrical Engineering, 2018, 66: 569-582.

[45]智宁, 毛善君, 李梅. 基于双伽马函数的煤矿井下低亮度图像增强算法[J]. 辽宁工程技术大学学报(自然科学版), 2018, 37(01): 191-197.

[46]Chang Y, Jung C, Ke P, et al. Automatic contrast-limited adaptive histogram equalization with dual gamma correction[J]. IEEE Access, 2018, 6: 11782-11792.

[47]Liu S, Wu H, Rahman M A, et al. Enhancement of low illumination images based on an optimal hyperbolic tangent profile[J]. Computers & Electrical Engineering, 2018, 70: 538-550.

[48]Huang Y, Li Y, Zhang Y. A Retinex image enhancement based on L channel illumination estimation and Gamma function[C]//2018 Joint International Advanced Engineering and Technology Research Conference (JIAET 2018). Atlantis Press, 2018: 312-317.

[49]David D. Low illumination image enhancement algorithm using iterative recursive filter and visual gamma transformation function[C]//2015 Fifth International Conference on Advances in Computing and Communications (ICACC). IEEE, 2015: 408-411.

[50]Wang W, Chen Z, Yuan X, et al. Adaptive image enhancement method for correcting low-illumination images[J]. Information Sciences, 2019, 496: 25-41.

[51]Hao S, Feng Z, Guo Y. Low-light image enhancement with a refined illumination map[J]. Multimedia Tools and Applications, 2018, 77(22): 29639-29650.

[52]Fu X, Zeng D, Huang Y, et al. A weighted variational model for simultaneous reflectance and illumination estimation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016: 2782-2790.

[53]Petro A B, Sbert C, Morel J M. Multiscale retinex[J]. Image Processing On Line, 2014: 71-88.

[54]Matin F, Jeong Y, Kim K, et al. Color image enhancement using multiscale Retinex based on particle swarm optimization method[C]//Journal of Physics: Conference Series. IOP Publishing, 2018, 960(1): 012026.

[55]Kim W. Image enhancement using patch-based principal energy analysis[J]. IEEE Access, 2018, 6: 72620-72628.

[56]Guo X, Li Y, Ling H. LIME: Low-light image enhancement via illumination map estimation[J]. IEEE Transactions on Image Processing, 2016, 26(2): 982-993.

[57]NASA, “Retinex image processing,” 2012. [Online]. Available:

https://dragon.larc.nasa.gov/retinex/pao/news

[58]Makhzani A, Frey B. K-sparse autoencoders[J]. arXiv preprint arXiv:1312.5663, 2013.

[59]Masci J, Meier U, Cireşan D, et al. Stacked convolutional auto-encoders for hierarchical feature extraction[C]//International conference on artificial neural networks. Springer, Berlin, Heidelberg, 2011: 52-59.

[60]Wang S, Zheng J, Hu H M, et al. Naturalness preserved enhancement algorithm for non-uniform illumination images[J]. IEEE Transactions on Image Processing, 2013, 22(9): 3538-3548.