改进的Retinex低照度图像光照补偿算法

摘要

针对Retinex算法在处理低照度图像时难以保证颜色恒常性以及边缘易出现光晕现象问题,提出了一种变换空间和引导滤波相结合的光照补偿算法.该算法将图像从RGB颜色空间转换到Lab颜色空间,用Retinex对L分量即亮度通道单独进行光照补偿处理,保持另外两个色彩通道a和b不变,将处理后的L通道和ab通道结合形成新的Lab′图像,然后将Lab′反转换到RGB空间得到最终补偿结果;在处理过程中,用引导滤波代替Retinex算法中的高斯滤波用于保持边缘特性和提高计算速率.最后通过直观结果、定量分析和处理时间三方面进行实验,实验结果表明,该光照补偿算法取得了较好的结果,能够保持较好的颜色恒常性、改善光晕现象以及减小时间复杂度.%To solve the problem that it is difficult to guarantee the color constancy and the edges tend to appear halo artifacts, a illumination compensation algorithm of space transformation with the combination of guided filter is proposed. First the RGB image is transformed into Lab space , Retinex algorithm is only used on the L channel to do illumination compensation , the other two channels a and b are kept steady, then the processed L channel and ab channels are merged into Lab', and the compensated image in Lab space is transformed back to RGB space , where gets the final image. During the processing, the Gaussian filter is substituted with a guided filter to reduce halo artifacts and improve computing rate. Finally through three ways including intuitionistic result , quantitative analysis and running time to conduct experiments for the proposed algorithm , the experimental results show that the proposed algorithm have achieved some good results , it can maintain good color constancy , reducing halo artifacts and improving operating efficiency.