搜索

x

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于明度分量的Retinex-Net图像增强改进方法

张航瑛 王雪琦 王华英 曹良才

引用本文:
Citation:

基于明度分量的Retinex-Net图像增强改进方法

张航瑛, 王雪琦, 王华英, 曹良才

Advanced Retinex-Net image enhancement method based on value component processing

Zhang Hang-Ying, Wang Xue-Qi, Wang Hua-Ying, Cao Liang-Cai
PDF
HTML
导出引用
  • 当人们在低照度光照条件下拍摄图像时, 图像通常会受到低可见度的影响. 这种低可见度的图像不仅影响视觉效果而且对后续的使用造成诸多困难. 为了解决低照度条件下图像可见度差, 色彩偏差等问题, 本文提出了一种改进的Retinex网络增强方法. 该方法首先对低照度RGB图像进行HSV色彩空间变换, 利用Retinex分解网络单独对明度分量进行分解增强, 并通过上采样操作增大明度分量的分辨率. 然后对色相分量和饱和度分量, 运用最近邻点插值增大其分辨率, 结合增强的明度分量转换回RGB色彩空间, 得到初始增强图像. 最后采用小波变换图像融合技术, 与原始低照度图像进行融合, 消除初始增强图像中的过度增强部分. 实验结果分析表明, 本文所提方法与原始Retinex网络方法相比, NIQE值平均下降了19.49%, 图像标准差平均提升了41.35%. 本文所提算法有望在安防监控、生物医学等领域得到有效应用.
    When capturing images under low-light lighting conditions, the resulting images often suffer low visibility. Such low-visibility images not only affect the visual effect but also cause many difficulties in practical application. Therefore, image enhancement technology under low-light conditions has always been a challenging problem in image algorithms. Considering that most of the existing image enhancement methods are based on the RGB color space enhancement technology, the correlation among the RGB three primary colors is ignored, which makes the color distortion phenomenon easy to occur when the image is enhanced. To solve the problems of poor image visibility and color deviation under low-light conditions, in this paper an advanced Retinex network enhancement method is proposed. In the method, firstly the low-light RGB image is transformed into HSV color space, the Retinex decomposition network is used to decompose and enhance the value component separately, and thus increasing the resolution of the value component through up-sampling operation; then, for the hue component and saturation component, the nearest neighbor interpolation is used to increase their resolutions, and the enhanced value component is combined to convert back to RGB color space to obtain the initial enhanced image; finally, the wavelet transform image fusion technology is used to fuse with the original low-light image to eliminate the over-enhanced part in the initial enhanced image. The analysis of experimental results shows that the method proposed in this paper has obvious advantages in brightness enhancement and color restoration of low-light images. Especially, comparing with the original Retinex network method, the NIQE value decreases by an average of 19.49%, and the image standard deviation increases by an average of 41.35%. The algorithm proposed in this paper is expected to be effectively used in the fields of security monitoring and biomedicine.
      通信作者: 曹良才, clc@tsinghua.edu.cn
    • 基金项目: 国家自然科学基金(批准号: 61827825)资助的课题.
      Corresponding author: Cao Liang-Cai, clc@tsinghua.edu.cn
    • Funds: Project supported by the National Natural Science Foundation of China (Grant No. 61827825 ).
    [1]

    蒋一纯, 詹伟达, 朱德鹏 2021 激光与光电子学进展 58 0410001Google Scholar

    Jiang Y C, Zhan W D, Zhu D P 2021 Laser Optoelectron. Prog. 58 0410001Google Scholar

    [2]

    韩平丽, 刘飞, 张广, 陶禹, 邵晓鹏 2018 物理学报 67 054202Google Scholar

    Han P L, Liu F, Zhang G, Tao Y, Shao X P 2018 Acta Phys. Sin. 67 054202Google Scholar

    [3]

    Liu J, Wang X, Chen M, Liu S G, Zhou X R, Shao Z F, Liu P 2014 Opt. Express 22 618Google Scholar

    [4]

    Fu X Y, Zeng D L, Huang Y, Liao Y H, Ding X H, Paisley J 2016 Signal Process. 129 82Google Scholar

    [5]

    Singh N, Bhandari A K 2021 IEEE Trans. Instrum. Meas. 70 1

    [6]

    Land E H 1964 Am. Sci. 52 247

    [7]

    Land E H, McCann J J 1971 J. Opt. Soc. Am. 61 1Google Scholar

    [8]

    Land E H, Hubel D H, Livingstone M S, Perry S H, Burns M M 1983 Nature 303 616Google Scholar

    [9]

    李红, 吴炜, 杨晓敏, 严斌宇, 刘凯, Gwanggil J 2016 物理学报 65 160701Google Scholar

    Li H, Wu W, Yang X M, Yan B Y, Liu K, Gwanggil J 2016 Acta Phys. Sin. 65 160701Google Scholar

    [10]

    Jobson D J, Rahman Z, Woodell G A 1997 IEEE Trans. Image Process. 6 451Google Scholar

    [11]

    Rahman Z, Jobson D J, Woodell G A 1996 Proceedings of 3rd IEEE International Conference on Image Processing Lausanne, Switzerland, September 19, 1996 p1003

    [12]

    Jobson D J, Rahman Z, Woodell G A 1997 IEEE Trans. Image Process. 6 965Google Scholar

    [13]

    毕国玲, 续志军, 赵建, 孙强 2015 物理学报 64 100701Google Scholar

    Bi G L, Xu Z J, Zhao J, Sun Q 2015 Acta Phys. Sin. 64 100701Google Scholar

    [14]

    Zhou Z Q, Dong M J, Xie X Z, Gao Z F 2016 Appl. Opt. 55 6480

    [15]

    王殿伟, 韩鹏飞, 范九 伦, 刘颖, 许志杰, 王晶 2018 物理学报 67 210701Google Scholar

    Wang D W, Han P F, Fan J L, Liu Y, Xu Z J, Wang J 2018 Acta Phys. Sin. 67 210701Google Scholar

    [16]

    Kwon H J, Lee S H, Lee G Y, Sohng K I 2014 Digit. Signal Process. 30 74Google Scholar

    [17]

    Yang Q X, Tan K H, Ahuja N 2009 IEEE Conference on Computer Vision and Pattern Recognition Miami, USA, June 20–25, 2009 p557

    [18]

    Wang S H, Zheng J, Hu H M, Li B 2013 IEEE Trans. Image Process. 22 3538Google Scholar

    [19]

    Fu X Y, Zeng D L, Huang Y, Zhang X P, Ding X H 2016 IEEE Conference on Computer Vision and Pattern Recognition Las Vegas, USA, June 27–30, 2016 p2782

    [20]

    Guo X J, Li Y, Ling H B 2017 IEEE Trans. Image Process. 26 982Google Scholar

    [21]

    Gijsenij A, Gevers T, Weijer J 2011 IEEE Trans. Image Process. 20 2475Google Scholar

    [22]

    赵欣慰, 金韬, 池灏, 曲嵩 2015 物理学报 64 104201Google Scholar

    Zhao X W, Jin T, Chi H, Qu S 2015 Acta Phys. Sin. 64 104201Google Scholar

    [23]

    Jiang Z Q, Li H T, Liu L j, Men A D, Wang H Y 2021 Neurocomputing 454 361Google Scholar

    [24]

    马红强, 马时平, 许悦雷, 朱明明 2019 光学学报 39 0210004Google Scholar

    Ma H Q, Ma S P, Xu Y L, Zhu M M 2019 Acta Opt. Sin. 39 0210004Google Scholar

    [25]

    Guo Y H, Ke X, Ma J, Zhang J 2019 IEEE Access 7 13737Google Scholar

    [26]

    Lore K G, Akintayo A, Sarkar S 2017 Pattern Recognit. 61 650Google Scholar

    [27]

    Wang W J, Wei C, Yang W H, Liu J Y 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition Xi'an, China, May 15–19, 2018 p751

    [28]

    He W J, Liu Y Y, Feng J F, Zhang W W, Gu G H, Chen Q 2020 IEEE 3rd International Conference on Information Systems and Computer Aided Education Dalian, China, September 27–29, 2020 p397

    [29]

    Wei C, Wang W J, Yang W H, Liu J Y 2018 arXiv: 1808.04560 v1 [cs. CV]

    [30]

    Yakno M, Mohamad-Saleh J, Ibrahim M Z 2021 Sensors 21 6445Google Scholar

    [31]

    陈刚, 刘言, 杨贺超, 孙斌, 喻春雨 2021 光学精密工程 29 1999Google Scholar

    Chen G, Liu Y, Yang H C, Sun B, Yu C Y 2021 Opt. Precis. Eng. 29 1999Google Scholar

    [32]

    Zhang H Y, Cao L C, Yang F 2021 Proc. SPIE First Optics Frontier Conference Hangzhou, China, June 18, 2021 1185002

    [33]

    Yadav A K, Roy R, Kumar A P, Kumar C S, Dhakad S K 2015 International Conference on Advances in Computing, Communications and Informatics Kochi, India, August 10–13, 2015 p1204

    [34]

    Mittal A, Soundararajan R, Bovik A C 2013 IEEE Signal Process. Lett. 20 209Google Scholar

  • 图 1  (a) Retinex成像理论模型; (b) Retinex算术思想简介

    Fig. 1.  (a) Retinex imaging theoretical model; (b) arithmetic ideas of Retinex algorithm.

    图 2  改进的Retinex网络增强算法流程图

    Fig. 2.  Flow chart of the advanced Retinex network enhancement algorithm.

    图 3  sym4尺度函数与小波函数

    Fig. 3.  Scale function and wavelet function of sym4.

    图 4  不同算法对比效果图

    Fig. 4.  Comparison of different algorithms.

    图 5  不同算法增强效果局部放大图

    Fig. 5.  Local enlarged view of the enhancement effect of different algorithms.

    图 6  不同方法图像评价指标的均值变化情况

    Fig. 6.  Changes in mean values of image evaluation metrics for different methods.

    表 1  V分量分解网络结构

    Table 1.  V-component decomposition network structure.

    输入操作卷积核输出通道步长输出
    RGBrgb to hsvH, S, V
    Vconv3641feats0
    feats0conv & ReLU3641feats1
    feats1conv & ReLU3641feats2
    feats2conv & ReLU3641feats3
    feats3conv & ReLU3641feats4
    feats4conv & ReLU3641feats5
    feats5conv & sigmoid321R, I
    下载: 导出CSV

    表 2  V分量增强网络结构

    Table 2.  V-component enhancement network structure.

    输入操作卷积核输出通道步长输出
    Vlow, Rlow, Ilowup-sampleInput
    Inputconv3641out0
    out0conv & ReLU3642out1
    out1conv & ReLU3642out2
    out2conv & ReLU3642out3
    out3interpolation64out3 up
    out3 up, out2
    de1
    conv & ReLU
    interpolation
    3
    64
    64
    1
    de1
    de1 up
    de1 up, out1
    de2
    conv & ReLU
    interpolation
    3
    64
    64
    1
    de2
    de2 up
    de2 up, out0de1
    de2
    conv & ReLUinterpolation
    interpolation
    3—
    6464
    64
    1—
    de3de1 r
    de2 r
    de1 r, de2 r, de3conv & ReLU3641feats0
    feats0conv1641feats1
    feats1conv311Vnew
    下载: 导出CSV

    表 3  不同图像的客观评价指标

    Table 3.  Objective evaluation metrics for different images.

    ImageEvaluateMSRCRAuto GCRetinex-NetARN
    Image1NIQE5.66925.13845.97824.0729
    Entropy7.10956.63927.13757.8179
    SD33.375841.695931.113042.9601
    Image 2NIQE6.29266.02525.35963.7336
    Entropy7.30127.51717.57777.7226
    SD41.890355.607146.342866.2911
    Image 3NIQE5.67154.92034.45284.0319
    Entropy6.78987.12247.72847.8633
    SD31.380037.302853.565472.4424
    Image 4NIQE3.76953.88443.72003.6582
    Entropy5.53927.18817.28077.4010
    SD40.891738.674139.991346.9674
    Image 5NIQE3.95414.47384.01263.6424
    Entropy7.34976.05497.28717.4387
    SD42.157441.086332.532156.5474
    Image 6NIQE7.34016.42735.44593.8790
    Entropy7.03355.57017.34177.8134
    SD34.113640.610838.497456.5800
    MeanNIQE5.44955.14494.76493.8363
    Entropy6.85386.68207.39227.6762
    SD37.301542.496240.300356.9647
    下载: 导出CSV
  • [1]

    蒋一纯, 詹伟达, 朱德鹏 2021 激光与光电子学进展 58 0410001Google Scholar

    Jiang Y C, Zhan W D, Zhu D P 2021 Laser Optoelectron. Prog. 58 0410001Google Scholar

    [2]

    韩平丽, 刘飞, 张广, 陶禹, 邵晓鹏 2018 物理学报 67 054202Google Scholar

    Han P L, Liu F, Zhang G, Tao Y, Shao X P 2018 Acta Phys. Sin. 67 054202Google Scholar

    [3]

    Liu J, Wang X, Chen M, Liu S G, Zhou X R, Shao Z F, Liu P 2014 Opt. Express 22 618Google Scholar

    [4]

    Fu X Y, Zeng D L, Huang Y, Liao Y H, Ding X H, Paisley J 2016 Signal Process. 129 82Google Scholar

    [5]

    Singh N, Bhandari A K 2021 IEEE Trans. Instrum. Meas. 70 1

    [6]

    Land E H 1964 Am. Sci. 52 247

    [7]

    Land E H, McCann J J 1971 J. Opt. Soc. Am. 61 1Google Scholar

    [8]

    Land E H, Hubel D H, Livingstone M S, Perry S H, Burns M M 1983 Nature 303 616Google Scholar

    [9]

    李红, 吴炜, 杨晓敏, 严斌宇, 刘凯, Gwanggil J 2016 物理学报 65 160701Google Scholar

    Li H, Wu W, Yang X M, Yan B Y, Liu K, Gwanggil J 2016 Acta Phys. Sin. 65 160701Google Scholar

    [10]

    Jobson D J, Rahman Z, Woodell G A 1997 IEEE Trans. Image Process. 6 451Google Scholar

    [11]

    Rahman Z, Jobson D J, Woodell G A 1996 Proceedings of 3rd IEEE International Conference on Image Processing Lausanne, Switzerland, September 19, 1996 p1003

    [12]

    Jobson D J, Rahman Z, Woodell G A 1997 IEEE Trans. Image Process. 6 965Google Scholar

    [13]

    毕国玲, 续志军, 赵建, 孙强 2015 物理学报 64 100701Google Scholar

    Bi G L, Xu Z J, Zhao J, Sun Q 2015 Acta Phys. Sin. 64 100701Google Scholar

    [14]

    Zhou Z Q, Dong M J, Xie X Z, Gao Z F 2016 Appl. Opt. 55 6480

    [15]

    王殿伟, 韩鹏飞, 范九 伦, 刘颖, 许志杰, 王晶 2018 物理学报 67 210701Google Scholar

    Wang D W, Han P F, Fan J L, Liu Y, Xu Z J, Wang J 2018 Acta Phys. Sin. 67 210701Google Scholar

    [16]

    Kwon H J, Lee S H, Lee G Y, Sohng K I 2014 Digit. Signal Process. 30 74Google Scholar

    [17]

    Yang Q X, Tan K H, Ahuja N 2009 IEEE Conference on Computer Vision and Pattern Recognition Miami, USA, June 20–25, 2009 p557

    [18]

    Wang S H, Zheng J, Hu H M, Li B 2013 IEEE Trans. Image Process. 22 3538Google Scholar

    [19]

    Fu X Y, Zeng D L, Huang Y, Zhang X P, Ding X H 2016 IEEE Conference on Computer Vision and Pattern Recognition Las Vegas, USA, June 27–30, 2016 p2782

    [20]

    Guo X J, Li Y, Ling H B 2017 IEEE Trans. Image Process. 26 982Google Scholar

    [21]

    Gijsenij A, Gevers T, Weijer J 2011 IEEE Trans. Image Process. 20 2475Google Scholar

    [22]

    赵欣慰, 金韬, 池灏, 曲嵩 2015 物理学报 64 104201Google Scholar

    Zhao X W, Jin T, Chi H, Qu S 2015 Acta Phys. Sin. 64 104201Google Scholar

    [23]

    Jiang Z Q, Li H T, Liu L j, Men A D, Wang H Y 2021 Neurocomputing 454 361Google Scholar

    [24]

    马红强, 马时平, 许悦雷, 朱明明 2019 光学学报 39 0210004Google Scholar

    Ma H Q, Ma S P, Xu Y L, Zhu M M 2019 Acta Opt. Sin. 39 0210004Google Scholar

    [25]

    Guo Y H, Ke X, Ma J, Zhang J 2019 IEEE Access 7 13737Google Scholar

    [26]

    Lore K G, Akintayo A, Sarkar S 2017 Pattern Recognit. 61 650Google Scholar

    [27]

    Wang W J, Wei C, Yang W H, Liu J Y 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition Xi'an, China, May 15–19, 2018 p751

    [28]

    He W J, Liu Y Y, Feng J F, Zhang W W, Gu G H, Chen Q 2020 IEEE 3rd International Conference on Information Systems and Computer Aided Education Dalian, China, September 27–29, 2020 p397

    [29]

    Wei C, Wang W J, Yang W H, Liu J Y 2018 arXiv: 1808.04560 v1 [cs. CV]

    [30]

    Yakno M, Mohamad-Saleh J, Ibrahim M Z 2021 Sensors 21 6445Google Scholar

    [31]

    陈刚, 刘言, 杨贺超, 孙斌, 喻春雨 2021 光学精密工程 29 1999Google Scholar

    Chen G, Liu Y, Yang H C, Sun B, Yu C Y 2021 Opt. Precis. Eng. 29 1999Google Scholar

    [32]

    Zhang H Y, Cao L C, Yang F 2021 Proc. SPIE First Optics Frontier Conference Hangzhou, China, June 18, 2021 1185002

    [33]

    Yadav A K, Roy R, Kumar A P, Kumar C S, Dhakad S K 2015 International Conference on Advances in Computing, Communications and Informatics Kochi, India, August 10–13, 2015 p1204

    [34]

    Mittal A, Soundararajan R, Bovik A C 2013 IEEE Signal Process. Lett. 20 209Google Scholar

  • [1] 施岳, 欧攀, 郑明, 邰含旭, 王玉红, 段若楠, 吴坚. 基于轻量残差复合增强收敛神经网络的粒子场计算层析成像伪影噪声抑制. 物理学报, 2024, 73(10): 104202. doi: 10.7498/aps.73.20231902
    [2] 刘鸿江, 刘逸飞, 谷付星. 基于深度学习的微纳光纤自动制备系统. 物理学报, 2024, 73(10): 104207. doi: 10.7498/aps.73.20240171
    [3] 战庆亮, 葛耀君, 白春锦. 基于深度学习的流场时程特征提取模型. 物理学报, 2022, 71(7): 074701. doi: 10.7498/aps.71.20211373
    [4] 张瑶, 张云波, 陈立. 基于深度学习的光学表面杂质检测. 物理学报, 2021, 70(16): 168702. doi: 10.7498/aps.70.20210403
    [5] 王甜甜, 王慧, 朱艳春, 王丽嘉. 基于位移流U-Net和变分自动编码器的心脏电影磁共振图像左心肌运动追踪. 物理学报, 2021, 70(22): 228701. doi: 10.7498/aps.70.20210885
    [6] 南虎, 麻晓晶, 赵海博, 汤少杰, 刘卫华, 王大威, 贾春林. 基于YOLOv3框架的高分辨电镜图像原子峰位置检测. 物理学报, 2021, 70(7): 076803. doi: 10.7498/aps.70.20201502
    [7] 苏博, 陶芬, 李可, 杜国浩, 张玲, 李中亮, 邓彪, 谢红兰, 肖体乔. 同步辐射纳米CT图像配准方法研究. 物理学报, 2021, 70(16): 160704. doi: 10.7498/aps.70.20210156
    [8] 周静, 张晓芳, 赵延庚. 一种基于图像融合和卷积神经网络的相位恢复方法. 物理学报, 2021, 70(5): 054201. doi: 10.7498/aps.70.20201362
    [9] 赵智鹏, 周双, 王兴元. 基于深度学习的新混沌信号及其在图像加密中的应用. 物理学报, 2021, 70(23): 230502. doi: 10.7498/aps.70.20210561
    [10] 郎利影, 陆佳磊, 于娜娜, 席思星, 王雪光, 张雷, 焦小雪. 基于深度学习的联合变换相关器光学图像加密系统去噪方法. 物理学报, 2020, 69(24): 244204. doi: 10.7498/aps.69.20200805
    [11] 陈炜, 郭媛, 敬世伟. 基于深度学习压缩感知与复合混沌系统的通用图像加密算法. 物理学报, 2020, 69(24): 240502. doi: 10.7498/aps.69.20201019
    [12] 刘杰, 张建勋, 代煜. 基于多引导滤波的图像增强算法. 物理学报, 2018, 67(23): 238701. doi: 10.7498/aps.67.20181425
    [13] 刘琦, 王丽丹, 段书凯. 一种基于忆阻交叉阵列的自适应三高斯模型及其在图像增强中的应用. 物理学报, 2017, 66(12): 127301. doi: 10.7498/aps.66.127301
    [14] 王聪, 杨晶, 潘秀娟, 蔡高航, 赵巍, 张景园, 崔大复, 彭钦军, 许祖彦. 基于光参量放大相位共轭特性的图像修复与增强. 物理学报, 2017, 66(10): 104205. doi: 10.7498/aps.66.104205
    [15] 李红, 吴炜, 杨晓敏, 严斌宇, 刘凯, Gwanggil Jeon. 基于主特征提取的Retinex多谱段图像增强. 物理学报, 2016, 65(16): 160701. doi: 10.7498/aps.65.160701
    [16] 毕国玲, 续志军, 赵建, 孙强. 基于照射_反射模型和有界运算的多谱段图像增强. 物理学报, 2015, 64(10): 100701. doi: 10.7498/aps.64.100701
    [17] 冯鑫, 李川, 胡开群. 基于深度玻尔兹曼模型的红外与可见光图像融合. 物理学报, 2014, 63(18): 184202. doi: 10.7498/aps.63.184202
    [18] 赵文达, 赵建, 续志军. 基于结构张量的变分多源图像融合. 物理学报, 2013, 62(21): 214204. doi: 10.7498/aps.62.214204
    [19] 甘甜, 冯少彤, 聂守平, 朱竹青. 基于分块DCT变换编码的小波域多幅图像融合算法. 物理学报, 2011, 60(11): 114205. doi: 10.7498/aps.60.114205
    [20] 张 闯, 柏连发, 张 毅. 基于灰度空间相关性的双谱微光图像融合方法. 物理学报, 2007, 56(6): 3227-3233. doi: 10.7498/aps.56.3227
计量
  • 文章访问数:  6624
  • PDF下载量:  145
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-01-14
  • 修回日期:  2022-02-09
  • 上网日期:  2022-03-04
  • 刊出日期:  2022-06-05

/

返回文章
返回