电子学报 ›› 2020, Vol. 48 ›› Issue (1): 180-188.DOI: 10.3969/j.issn.0372-2112.2020.01.022

• 学术论文 • 上一篇    下一篇

基于卷积分析稀疏表示和相位一致性的低照度图像增强

周浦城, 张杰, 薛模根, 尹璋堃   

  1. 陆军炮兵防空兵学院信息工程系, 安徽合肥 230031
  • 收稿日期:2018-07-21 修回日期:2019-03-03 出版日期:2020-01-25 发布日期:2020-01-25
  • 作者简介:周浦城 男,1977年9月出生于江西宜春.2006年毕业于哈尔滨工业大学计算机应用技术专业获工学博士学位.现为陆军炮兵防空兵学院信息工程系副教授、硕士生导师.主要研究方向为图像处理与分析、信息融合技术.E-mail:zhoupc@hit.edu.cn;张杰 男,1994年2月出生于河北张家口.硕士,主要研究方向为数字图像处理.E-mail:zj199402@163.com
  • 基金资助:
    国家自然科学基金(No.61379105);安徽省自然科学基金(No.1908085MF208)

Low-light Image Enhancement Based on Convolutional Analysis Sparse Representation and Phase Congruency

ZHOU Pu-cheng, ZHANG Jie, XUE Mo-gen, YIN Zhang-kun   

  1. Department of Information Engineering, Army Academy of Artillery and Air Defence Forces, Hefei, Anhui 230031, China
  • Received:2018-07-21 Revised:2019-03-03 Online:2020-01-25 Published:2020-01-25

摘要: 针对低照度图像存在的对比度低、视觉效果差等问题,提出一种基于卷积分析稀疏表示和相位一致性的低照度图像增强方法.该方法基于Retinex模型,在估计照度图像时采用卷积分析稀疏表示进行约束,所用滤波器一部分人工设定,一部分由样本训练自动获得;在计算反射图像时利用单演相位一致性特征,施加相位一致性残余最小约束来恢复细节;通过联合约束并进行优化,得到的反射图像即为最终的增强结果.对大量低照度图像进行实验,并与当前先进方法相比,结果表明,本文方法不仅提高了图像的亮度与对比度,增强了细节,而且在多个客观评价指标上都优于其他方法.

关键词: 低照度图像, Retinex模型, 卷积分析稀疏表示, 相位一致性

Abstract: Low-light images suffer from low visibility and poor visual quality.To improve the quality of low-light images,a method based on convolution analysis sparse representation and phase congruency was proposed.This method is based on the Retinex model and improves the problem of insufficient constraints.More concretely,we used the convolutional analysis sparse representation whose filters were hand-crafted and learned from the input to estimate the illumination image.Then,by using the monogenic phase congruency,the reflection image was calculated via the phase congruency residual minimization to enhance weak details.Through joint constraints and optimization,the resulting reflection image served as the final enhancement result.Experiments on a number of challenging low-light images are presented to reveal the efficacy of our method and show its superiority over several state-of-the-arts on both subjective and objective assessments.

Key words: low-light image, Retinex model, convolutional analysis sparse representation, phase congruency

中图分类号: