图像在采集过程中，会受到天气、光照强度、光源方向等多种复杂因素的干扰，导致图像出现对比度差、亮度过低或过高、细节丢失、色彩失真、含有大量噪声等问题，影响对图像中信息的有效获取。因此，研究复杂光照环境下的图像增强算法具有十分重要的意义。根据不同的应用需求，论文分别提出了一种基于多尺度梯度域引导滤波的自适应Retinex图像增强算法和一种基于注意力机制与U-Net++网络的图像增强算法。具体研究内容如下：

       针对经典Retinex图像增强算法在提升亮度和对比度的同时，往往会导致细节丢失、光晕伪影出现、噪声放大及色彩失真等问题，论文对现有的Retinex增强算法进行了改进，提出了一种基于多尺度梯度域引导滤波的自适应Retinex图像增强算法。首先，设计了一种自适应伽马校正函数，能够在提升暗区域亮度的同时，有效抑制亮区域过增强，从而避免了亮区域细节丢失的问题；其次，采用多尺度梯度域引导滤波作为中心环绕函数，由于其具备各向异性的特点，从而能够准确估计光照分量，有效避免光晕伪影，同时保留边缘细节；然后，分别采用梯度域引导滤波和多尺度细节提升对反射分量进行处理，从而避免增强过程中的噪声放大问题并充分保留了纹理细节；最后，通过增强后的亮度图像计算亮度增益矩阵，实现对RGB图像逐像素的颜色恢复，从而有效避免色彩失真。实验结果表明，从主客观角度与现有算法对比，经该算法处理后图像亮度和对比度提高比例适中，噪声有效去除，细节保留完整，色彩更加自然、生动，视觉感官效果更佳，同时执行效率较高。

       针对Retinex及其改进的图像增强算法对极端光照环境下图像增强效果有限，易出现细节模糊、色彩不饱和、噪声抑制能力不佳及过增强等问题，论文提出了一种基于注意力机制与U-Net++网络的图像增强算法。首先，采用具有全方位学习特征能力的U-Net++构建图像分解网络，准确分解出图像的精细特征，从而有效避免分解过程中的细节模糊问题；其次，联合注意力机制和具有去噪能力的U-Net设计反射分量精细化恢复网络，使网络在去噪过程中重点关注纹理等有利信息，从而有效抑制噪声，恢复反射分量的纹理细节，并在此基础上引入色彩饱和度损失函数，使反射分量尽可能保留原有色彩；最后，引入自适应亮度调节比例图，并结合特征复用和相对位置的思想，重复使用相同大小的卷积核构建光照分量调节网络，从而有效避免过增强问题，保持图像细节信息。实验结果表明，在多种数据集上与主流算法相比，该算法主客观评价综合最优，对多种极端光照环境下图像均能产生较佳的增强效果，细节清晰，色彩真实自然，同时能够有效抑制噪声和过增强。

    Images are interfered by various complex factors such as weather, light intensity and light source direction during the acquisition process, resulting in poor contrast, too low or too high brightness, loss of details, color distortion and a lot of noise, which seriously affects the effective acquisition of useful information in images. Therefore, it is of great importance to study the image enhancement algorithm under complex lighting environment. In this paper,  an adaptive Retinex image enhancement algorithm based on multi-scale gradient domain guided filtering and a image enhancement algorithm based on attention mechanism and U-Net++ network are proposed respectively according to different application scenarios. The details of the research are as follows.

    Aiming at the problems that the classical Retinex image enhancement algorithm often leads to loss of details, appearance of halo artifacts, noise amplification and color distortion while enhancing brightness and contrast, this paper improves the existing Retinex enhancement algorithm and puts forward an adaptive Retinex image enhancement algorithm based on multiscale gradient domain guided filtering. First, an adaptive gamma correction function is designed, which can effectively suppress the over-enhancement of bright regions while enhancing the luminance of dark regions, thus avoiding the problem of detail loss in bright regions; second, a gradient domain guided image filter with anisotropic properties is used as the central surround function, which is combined with a multiscale strategy to accurately estimate the illumination components, thus effectively avoiding halo artifacts while preserving the edge details; then, the gradient domain guided image filter and multiscale detail enhancement are used to process the reflection components respectively, so as to avoid the noise amplification problem in the enhancement process and fully retain the texture details; finally, the brightness gain matrix from the enhanced luminance image is calculated to recover the RGB image pixel by pixel, so as to effectively avoid the color distortion. The experimental results show that, compared with the existing algorithm from both subjective and objective perspectives, the image brightness and contrast are improved moderately, noise is effectively removed, details are retained intact, colors are more natural and vivid, and the visual sensory effect is better, while the execution efficiency is higher.

    In response to Retinex and its improved enhancement algorithms's limited effect on image enhancement under extreme illumination, which is prone to blurred details, unsaturated colors, poor noise suppression and over-enhancement, this paper proposes an image enhancement algorithm based on attention mechanism and U-Net++ network. First, U-Net++, which has the ability to learn features in all directions, is used to build an image decomposition network to accurately decompose the image, thus effectively avoiding the blurred details in the decomposition process; second, the reflection component recovery network is designed by combining the attention mechanism and U-Net with denoising capability, so that the network focuses on favorable information such as color and edge in the recovery process to effectively suppress noise and recover the reflection component texture details, and the color saturation loss function, on this basis, is introduced to make the reflection component retain the original color as much as possible; finally, the adaptive luminance adjustment scale map is introduced and combined with the ideas of feature reuse and relative position to build the light component adjustment network, so as to effectively avoid the over-enhancement problem while maintaining the image detail information. Compared with the mainstream algorithms on a variety of datasets, the experimental results show that the algorithm has the best comprehensive subjective and objective evaluation, and can produce a better enhancement effect for images under various extreme illumination, with clear details and true colors, and can effectively suppress noise and over-enhancement.

[1]Zamir S W, Arora A, Khan S, Hayat M, Khan F, Yang M H, Shao L. Multi-stage progressive image restoration[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021: 14821-14831.

[2]Saleh K, Szenasi S, Vamossy Z. Occlusion Handling in Generic Object Detection: A Review[C]//2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics. 2021: 477-484.

[3]Mai Z, Li R, Jeong J, Quispe D, Kim H, Sanner S. Online continual learning in image classification: An empirical survey[J]. Neurocomputing, 2022, 469: 28-51.

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

[5]朱铮涛, 萧达安. 基于非线性调整的伽马校正图像增强算法[J]. 计算机工程与设计, 2018, 39(9): 2822-2826.

[6]刘志成, 王殿伟, 刘颖, 刘学杰. 基于二维伽马函数的光照不均匀图像自适应校正算法[J]. 北京理工大学学报, 2016, 36(2): 191-196.

[7]Huang Z, Fang H, Li Q, Li Z, Zhang T, Sang N, Li Y. Optical remote sensing image enhancement with weak structure preservation via spatially adaptive gamma correction[J]. Infrared Physics & Technology, 2018, 94: 38-47.

[8]Shin Y, Jeong S, Lee S. Content awareness-based color image enhancement[C]//2014 International Symposium on Consumer Electronics. 2014: 1-2.

[9]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(9): 1830-1835

[10]汪子玉,朱益铭,戴声奎.基于图像序列分析的全局直方图均衡[J].信号处理, 2020, 36(4): 531-540

[11]Yelmanov S, Hranovska O, Romanyshyn Y. A new approach to the implementation of histogram equalization in image processing[C]//2019 3rd International Conference on Advanced Information and Communications Technologies. 2019: 288-293.

[12]Nithyananda C R, Ramachandra A C. Review on histogram equalization based image enhancement techniques[C]//2016 International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT). 2016: 2512-2517.

[13]Shi Z, Feng Y, Zhao M, Zhang E, He L. Normalised gamma transformation-based contrast-limited adaptive histogram equalisation with colour correction for sand-dust image enhancement[J]. IET Image Process, 2020, 14(4): 747-756.

[14]杨嘉能, 李华, 田宸玮, 钱育蓉. 基于自适应校正的动态直方图均衡算法[J]. 计算机工程与设计, 2021, 42(5): 1264-1270.

[15]Wang Y, Pan Z. Image contrast enhancement using adjacent-blocks-based modification for local histogram equalization[J]. Infrared Physics & Technology, 2017, 86: 59-65.

[16]Pizer S, Austin J. Adaptive histogram equalization and its variations [J]. Computer Vision Graphics & Image Processing, 1987, 39(3): 355-368.

[17]Zuiderveld K. Contrast Limited Adaptive Histogram Equalization[J]. Graphics Gems, 1994, 5:474-485.

[18]Singh K, Kapoor R. Image enhancement using Exposure based Sub Image Histogram Equalization[J]. Pattern Recognition Letters, 2014, 36(15): 10-14.

[19]Somal S. Image enhancement using local and global histogram equalization technique and their comparison[C]//First International Conference on Sustainable Technologies for Computational Intelligence. 2020: 739-753.

[20]Land E H. The Retinex Theory of Color Vision[J]. Scientific American, 1978, 237(6): 108-128.

[21]Jobson D, Rahman Z. Properties and performance of a center/surround retinex[J]. IEEE Transactions on Image Processing, 1997, 6(3): 451-462.

[22]Jobson D, Rahman Z. A multiscale retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image processing, 1997, 6(7): 965-976

[23]Rahman Z, Jobson D. Retinex processing for automatic image enhancement[C] //Human Vision and Electronic Imaging VII. International Society for Optics and Photonics. 2004: 100-110.

[24]Yao L, Lin Y, Muhammad S. An Improved Multi-Scale Image Enhancement Method Based on Retinex Theory[J]. Journal of Medical Imaging and Health Informatics, 2018, 8(1): 122-126.

[25]张杰, 周浦城, 张谦. 基于迭代多尺度引导滤波Retinex的低照度图像增强[J]. 图学学报, 2018, 39(1): 1-11.

[26]Ghosh S, Chaudhury K. Fast bright-pass bilateral filtering for low-light enhancement[C]// 2019 IEEE International Conference on Image Processing. 2019: 205-209.

[27]Ma J, Fan X, Ni J, Zhu X, Xiong C. Multi-scale retinex with color restoration image enhancement based on Gaussian filtering and guided filtering[J]. International Journal of Modern Physics B, 2017, 31: 16-19.

[28]Li H, Wang R, Geng Z, Hu H. Low-illumination image enhancement algorithm based on multi-scale gradient domain guided filtering[J]. Journal of Computer Applications, 2019, 39(10): 3046-3052.

[29]韩梦妍, 李良荣, 蒋凯. 基于光照图估计的Retinex低照度图像增强算法[J]. 计算机工程, 2021, 47(10): 201-206.

[30]Fu X, Liao Y, Zeng D, Huang Y, Ding X. A Probabilistic Method for Image Enhancement With Simultaneous Illumination and Reflectance Estimation[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2015, 24(12): 4965-4977.

[31]胡韦伟, 汪荣贵, 方帅, 胡琼. 基于双边滤波的Retinex图像增强算法[J].工程图学学报, 2010, 31(2): 104-109.

[32]Yin J, Li H, Du J, He P. Low illumination image Retinex enhancement algorithm based on guided filtering[C]//2014 IEEE 3rd International Conference on Cloud Computing and Intelligence Systems. 2014: 639-644.

[33]Sun X, Liu H, Wu S, Fang Z, Li C, Yin J. Low-Light Image Enhancement Based on Guided Image Filtering in Gradient Domain[J]. International Journal of Digital Multimedia Broadcasting, 2017(2): 1-13.

[34]Wang S, Zheng J, Hu H. 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): 3538-3548.

[35]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.

[36]Li M, Liu J, Yang W, Sun X, Guo Z. Structure-Revealing Low-Light Image Enhancement Via Robust Retinex Model[J]. IEEE Transactions on Image Processing, 2018, 27(6): 2828-2841.

[37]Asmare M, Asirvadam V, Hani A. Image enhancement based on contourlet transform[J]. Signal Image and Video Processing, 2015, 9(7): 1679-1690.

[38]Kamiyama M, Taguchi A. HSI color space with same gamut of RGB color space[J]. IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences, 2017, 100(1): 341-344.

[39]Zhou M, Jin K, Wang S, Ye J, Qian D. Color retinal image enhancement based on luminosity and contrast adjustment[J]. IEEE Transactions on Biomedical Engineering, 2018, 65(3): 521-527.

[40]王殿伟, 王晶, 许志杰, 刘颖. 一种光照不均匀图像的自适应校正算法[J]. 系统工程与电子技术, 2017, 39(6): 1383-1390.

[41]田会娟, 蔡敏鹏, 关涛, 胡阳. 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, 49(02): 173-184.

[42]Gu J, Wang Z, Kuen J, Ma L, Shahroudy A, Shuai B, Liu T, Wang X, Wang G, Cai J, Chen T. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377.

[43]Lore K, Akintayo A, Sarkar S. LLNet: A Deep Autoencoder Approach to Natural Low-light Image Enhancement[J]. Pattern Recognition, 2017, 61: 650-662.

[44]Wang W, Wei C, Yang W, Liu J. Gladnet: Low-light enhancement network with global awareness[C]//2018 13th IEEE international conference on automatic face & gesture recognition. 2018: 751-755.

[45]Shen L, Yue Z, Feng F, Chen Q, Liu S, Ma J. Msr-net: Low-light image enhancement using deep convolutional network[J]. arXiv preprint arXiv:1711.02488, 2017.

[46]Wei C, Liu J, Yang W. Deep retinex decomposition for low-light enhancement[J]. arXiv preprint arXiv:1808.04560, 2018.

[47]Dabov K, Foi A, Katkovnik V, Egiazarian K. Image denoising with block-matching and 3D filtering[C]//Image Processing: Algorithms and Systems, Neural Networks, and Machine Learning. 2006: 606414.1-606414.12.

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

[49]Guo C, Li C, Guo J, Loy C, Hou J, Kwong S, Cong R. Zero-reference deep curve estimation for low-light image enhancement[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 1780-1789.

[50]谢凤英, 汤萌, 张蕊. 基于Retinex的图像增强方法综述[J]. 数据采集与处理, 2019, 34(1):1-11.

[51]Foster D H . Color constancy[J]. Vision Research, 2011, 51( 7): 674-700.

[52]Akiyoshi K, Baingio P. Orientation misperceptions induced by contrast polarity:Comment on Contrast polarities determine the direction of Café Wall tilts[J]. Perception, 2006, 35(3): 401-409.

[53]吴文灿. 关于色觉理论的一种新见解[J]. 生理科学, 1988(01): 61-65.

[54]Balaraman S. Color vision research and trichromatic theory: A historical review[J]. Psychological bulletin, 1962, 59(5): 434-448

[55]Niall K. On the Relation of the Tetrachromatic Theory to the Trichromatic Theory[M]. Canada: Cham, 2017: 167-185.

[56]李淼, 周冬明, 刘琰煜, 谢诗冬, 王长城, 卫依雪. 结合深度残差神经网络与Retinex理论的低照度图像增强[J]. 云南大学学报(自然科学版), 2021, 43(4): 690-699.

[57]牟琦, 魏妍妍, 李姣, 李洪安, 李占利. 改进的Retinex低照度图像增强算法研究[J]. 哈尔滨工程大学学报, 2018, 39(12): 2001-2010.

[58]Tomasi C, Manduchi R. Bilateral filtering for gray and color images[C]//Sixth international conference on computer vision. 1998: 839-846.

[59]He K, Sun J, Tang X. Guided image filtering[J]. IEEE transactions on pattern analysis and machine intelligence, 2012, 35(6): 1397-1409.

[60]Zhang W, Tanida J, Itoh K, Ichioka Y. Shift-invariant pattern recognition neural network and its optical architecture[C]//Proceedings of annual conference of the Japan Society of Applied Physics. 1988: 2147-2151.

[61]Le C, Bengio Y, Hinton G. Deep learning[J]. nature, 2015, 521(7553): 436-444.

[62]Le C, Boser Y, Denker B, Hender D. Back-propagation applied to handwritten zip-code recognition[J]. Neural computation, 1989, 1(4): 541-551.

[63]Han J, Moraga C. The influence of the sigmoid function parameters on the speed of backpropagation learning[C]//International workshop on artificial neural networks. 1995: 195-201.

[64]Nair V, Hinton G. Rectified linear units improve restricted boltzmann machines[C]//Icml. 2010: 807-814.

[65]Maas A L, Hannun A Y, Ng A Y. Rectifier nonlinearities improve neural network acoustic models[C]//Proc Int conf on Machine Learning. 2013: 1-3.

[66]Long J, Shelhamer E, Darrell T. Fully Convolutional Networks for Semantic Segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(4): 640-651.

[67]Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention. 2015: 234-241.

[68]Zhou Z, Siddiquee M, Tajbakhsh N, Liang J. Unet++: A nested u-net architecture for medical image segmentation[M]//Deep learning in medical image analysis and multimodal learning for clinical decision support. Switzerland: Springer, 2018: 3-11.

[69]Jaderberg M, Simonyan K, Zisserman A. Spatial transformer networks[J]. Advances in neural information processing systems, 2015, 28-37.

[70]Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 7132-7141.

[71]Wang F, Jiang M, Qian C, Yang S, Li C, Zhang H, Wang X, Tang X. Residual attention network for image classification[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 3156-3164.

[72]Fu J, Liu J, Tian H, Li Y, Bao Y, Fang Z, Lu H. Dual attention network for scene segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 3146-3154.

[73]Woo S, Park J, Lee J Y, Kweon S. Cbam: Convolutional block attention module[C]// Proceedings of the European conference on computer vision. 2018: 3-19.

[74]Niu Z, Zhong G, Yu H. A review on the attention mechanism of deep learning[J]. Neurocomputing, 2021, 452: 48-62.

[75]Kou F, Chen W, Wen C, Li Z. Gradient domain guided image filtering[J]. IEEE Transactions on Image Processing, 2015, 24(11): 4528-4539.

[76]Shin Y, Jeong S, Lee S. Content awareness-based color image enhancement[C] //The 18th IEEE International Symposium on Consumer Electronics (ISCE). 2014: 1-2.

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

[78]Gorai A, Ghosh A. Hue-preserving color image enhancement using particle swarm optimization[C]. Recent Advances in Intelligent Computational Systems. 2011: 563-568

[79]Dong X, Wang G, Pang Y, Li W, Wen J, Meng W, Lu Y. Fast efficient algorithm for enhancement of low lighting video[C]//2011 IEEE International Conference on Multimedia and Expo. 2011: 1-6.

[80]Fu X, Zeng D, Huang Y, Liao Y, Ding X, Paisley J. A fusion-based enhancing method for weakly illuminated images[J]. Signal Processing, 2016, 129: 82-96.

[81]Lee C, Kim C. Contrast enhancement based on layered difference representation [C]//2012 19th IEEE international conference on image processing. 2012: 965-968.

[82]Ignatov A, Kobyshev N, Timofte R, Vanhoey K, Gool L. Dslr-quality photos on mobile devices with deep convolutional networks[C]//Proceedings of the IEEE International Conference on Computer Vision. 2017: 3277-3285.

[83]Vonikakis V, Kouskouridas R, Gasteratos A. On the evaluation of illumination compensation algorithms[J]. Multimedia Tools and Applications, 2018, 77(8): 9211-9231.

[84]Boer J, Cense B, Park B, Piere M C, Tearney G, Bouma B. Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography[J]. Optics Letters, 2003, 28(21): 2067-2069.

[85]Wang Z, Bovik A C, Sheikh H R, Simoncelli E P. Image quality assessment: from error visibility to structural similarity[J]. IEEE transactions on image processing, 2004, 13(4): 600-612.

[86]Mittal A, Soundararajan R, Bovik A. Making a Completely Blind Image Quality Analyzer[J]. IEEE Signal Processing Letters, 2013, 20(3): 209-212.