Deep Unfolded Underwater Image Enhancement Based on Extreme Channels Prior

We propose a deep unrolling approach for underwater image enhancement using extreme channels prior. First, we formulate underwater image enhancement as a joint optimization problem that incorporates an underwater-related extreme channels prior and implicit regularization functions. Then, we solve the optimization problem iteratively and develop an unfolded deep neural network, where each block of the network represents an iteration in which the optimization variables and regularizers are updated using closed-form solutions and learned proximal operators, respectively. Experimental results demonstrate that the proposed algorithm outperforms state-of-the-art underwater image enhancement algorithms in both quantitative and qualitative comparisons.