液晶与显示, 2017, 32 (10): 835, 网络出版: 2017-11-27
基于正交偏振方向的多幅图像复原新方法
New method of multiple image restoration based on orthogonal polarization direction
偏振光 双边滤波 大气耗散函数 图像评价指标 晕轮效应 polarized light bilateral filtering atmospheric veil image evaluation index halo effect
摘要
本文根据大气场景成像过程中的偏振效应原理, 建立偏振成像方程, 提出一种基于两幅偏振化方向互相垂直的图像复原新方法。该方法从多幅偏振图像复原原理出发, 结合大气散射物理模型推导出新的偏振图像复原方程; 采用双边滤波原理估算偏振图像大气耗散函数; 结合图像评价指标自适应调整大气耗散函数, 寻找最佳雾浓度分布; 选取合适参数调整方法解决含有大面积天空区域雾霾图像复原所产生的晕轮效应。对采用传感器和KENKO CPL67MM偏振滤镜所获取的偏振图像进行大量实验测试, 并和当前较具代表性的单幅图像复原方法进行比较分析。结果表明, 本文的多幅偏振图像复原方法取得了更好的图像复原效果。
Abstract
According to the principle of polarization effect in the process of atmospheric scene imaging, the polarization imaging equation is established, and a new image restoration method based on the two polarization directions is proposed. A new polarization image restoration equation is derived by combining the atmospheric scattering physical model based on the principle of multiple polarization image restoration. First, the method uses bilateral filtering to estimate the atmospheric veil by the polarization image, and adaptively adjust it according to the image evaluation index. Finally, we restore the original clear image of polarization image combining the image matching and fusion restoration equation and select the appropriate parameter adjustment method to solve the large sky area containing the halo effect caused by haze image restoration. The experimental results by the polarization images obtained by the sensor and the KENKO CPL67MM polarizing filter were compared with those of the representative single image restoration methods. The results show that the algorithm can effectively recover the fog image and automatically adjust the parameters because the restoration information is more abundant.
彭文竹, 张禹, 王钦, 吴亚建. 基于正交偏振方向的多幅图像复原新方法[J]. 液晶与显示, 2017, 32(10): 835. PENG Wen-zhu, ZHANG Yu, WANG Qin, WU Ya-jian. New method of multiple image restoration based on orthogonal polarization direction[J]. Chinese Journal of Liquid Crystals and Displays, 2017, 32(10): 835.