Light Attenuation and Color Fluctuation for Underwater Image Restoration

Jingchun Zhou, Dingshuo Liu, Dehuan Zhang, Weishi Zhang; Proceedings of the Asian Conference on Computer Vision (ACCV), 2022, pp. 3845-3860

Underwater images with low contrast and color distortion of different scenarios often pose a significant challenge in improving image quality. Therefore, guidance by estimates of degenerated parameters related to its variety is beneficial, and they should be updated with degraded scenes to allow optimal vision. We propose a robust underwater image restoration method. Specifically, we adjust the color of the input image according to a background light estimation strategy guided by the depth map, light absorption priors considering hue, and Information entropy. At the same timewe adjust the depth map of the input image by computing color fluctuation and falloff. According to qualitative and quantitative evaluation, the proposed method generates vivid results with a more natural appearance and more valuable information.