Underwater Image Enhancement via Piecewise Colour Balancing and Multiscale Enhancement Fusion

Underwater images often suffer from multiple degradation issues, such as color casting, low contrast levels, and blurry details, which limits the applicability of underwater images in ocean exploration tasks. An underwater image enhancement method implemented via piecewise color balancing and pyramid-based contrast enhancement (PBPE) is proposed in this article. Concretely, PBPE first uses a reference channel with the maximum mean and two gain factors to balance the differences among the r, g, and b channels. PBPE presents a pixelwise transmission estimation method that is based on the mapping between the transmission and the backscattered light. Specifically, an adaptive compensation strategy is proposed to adaptively refine the transmission. Finally, PBPE captures a detail pyramid via multiscale Gaussian decomposition and uses the estimated transmission to remove haze and increase the degree of detail, thereby enhancing the overall contrast level of the underwater image. Comprehensive experiments conducted on two underwater image enhancement data sets indicate that our PBPE approach achieves better results and outperforms the state-of-the-art methods; i.e., compared with those of the second-best method, the average blur and density of the fog assessment-based defogger values of our method decrease by at least 2.67% and 4.55%, respectively, which shows that our method achieves enhancement results with high contrast levels and natural appearances.