Uncertainly Guided Multi-Scale Underwater Image Enhancement Network with Tone Mapping Block

Underwater image enhancement remains a challenging problem for severe color distortions of the underwater images. In this work, an Uncertainly Guided Multi-scale Underwater Image Enhancement Network (UMUEN) with Tone Mapping Block is introduced by us to address this problem. In the proposed framework, a confidence module is induced and the confidence map is utilized as guidance to reconstruct more confident enhanced results in training step. So as to resolve the color distortion problems, we further design a tone mapping block to correct the color shift. The effectiveness of different constituents of the UMUEN is showed through an ablation study. Large quantities of experiment results demonstrate that the proposed UMUEN performs better compare with the state-of-the-art methods on real-world datasets.