Efficient Image Dehazing by Improving Retinex Algorithm
Juntao Xue, Junjie Hei, Kaiyu Li, Ruohan Ma · 2020
The traditional Retinex hazy image enhancement algorithm using a fixed-scale parameter filter without considering the disadvantages of hazy density, so we propose an improved Retinex dehazing algorithm based on the dark channel prior. According to the correspondence between filter scale parameters and mean values of image sub-block, this algorithm uses Gaussian filters with different scale parameters for different haze concentration regions. Secondly, this algorithm enhanced image performance with the help of Retinex. Finally, optimized details by restoring color information and enhancing image edges with improved fractional differentiation. Experiments prove that this algorithm improves the sharpness and contrast of the image, makes the image richer in detail, has a stronger edge retention ability, and also restores the image with better layering and global effects.