Multiscale Fusion Algorithm for Underwater Image Enhancement Based on Color Preservation
Ming Yin, Xiaoxuan Du, Wei Liu, Liping Yu, Yan Xing
Abstract
Underwater images regularly exhibit color deviation and low contrast attributed to attenuation and scattering associated with wavelength and distance. For solving these two degradation problems, we propose an efficient underwater image enhancement model, which takes a multiscale fusion strategy as a pivot to concurrently achieve color balance and contrast enhancement. Specifically, we target two core degradation problems through two preprocessing steps. On the one hand, the white balance strategy enhances the appearance of the image by suppressing unnecessary colors to solve the color deviation problem. On the other hand, we employ a color-preserving adaptive histogram equalization (CP-AHE) to obtain an image with higher contrast by jointly processing the three color channels. In the fusion phase, a fusion strategy based on the detail-preserving decomposition of the edge-protected structural blocks is used to converge the white balance and CP-AHE inputs. Extensive experiments on common underwater image datasets have shown the advantages of the proposed method over other state-of-the-art underwater image methods in terms of visual and quantitative evaluation.