Underwater Image Enhancement based on Light Attenuation Prior

Abstract

Due to light dispersion and absorption in the underwater medium, images taken underwater often have poor contrast, color distortion, detail blurring, and a blue or greenish tone, all of which negatively impact visibility. Restoration approaches and enhancement methods are the two main categories into which underwater image processing technologies fall. In restoration techniques, the impacts of the underwater environment are thought to be degrading; however, in enhancement, this environment is assumed to be natural and attempts to further improve the visual information. It has been shown that restoration strategies perform better than augmentation programs. These restoration techniques are further divided into three categories: previous knowledge, polarization, and optical imaging. These schemes' main issues include overexposure, red artifacts, difficulties identifying artificial illumination, responding to multi-scatter scenarios, and unnecessary optimization parameters. Many systems fall into these categories in the literature, and the majority of them have one or more of the aforementioned problems. This research will offer a quick and efficient scene depth estimation model based on underwater light attenuation prior for underwater photos. Learning-based supervised linear regression will be used to train the model coefficients. The real scene radiance underwater may be readily recovered by estimating the background light and transmission maps for R-G-B light using the appropriate depth map.