Enhancing underwater images with medium-guided multi-color space embedding

Abstract

Our strategy makes use of the unique characteristics of the underwater medium to direct the enhancing procedure.Using a scattering and absorption model based on physics, we first simulate the color distortion brought on by the water medium.This model describes how underwater circumstances affect color appearance and picture quality.We build a medium transmission (showing the fraction of scene light reaching the camera)guided decoder network inspired by underwater imaging physical models to improve network performance in low-quality locations.As a consequence, our network can successfully improve the visual quality of underwater photographs by combining the advantages of physical model-based and learning-based techniques.Extensive testing shows that our Ucolor outperforms cutting-edge approaches in terms of visual quality and quantitative factors.