Multi-Indicator reconstruction for underwater polarized image dehazing method

Aiming at the limitations of traditional active polarization imaging methods that use a single evaluation indicator to invert polarization parameters and the insufficient recovery of underwater turbidity recovery details, this study proposes a multi-indicator reconstruction polarization image clarification method. We consider a variety of evaluation indicators during the imaging process to further enhance the clarity quality of underwater polarization images. First, the two optimal images for the evaluation indicators Enhancement Measure Evaluation (EME) and Mutual Information (MI) between the background light and the target light are inverted using an active polarization imaging model. Then, the detail layer of the EME-optimal image and the structural layer of the MI-optimal image are decomposed. Finally, we use the detail layer of the EME-optimal image and the structural layer of the MI-optimal image for weighted fusion, and find the optimal weights by genetic algorithm using two evaluation indicators, contrast and information entropy, as the fitness function. The experimental results show that the imaging quality of this method in turbid media with different concentrations and different objects with high and low polarization is significantly improved as measured by a variety of objective evaluation indicators. It can enhance the perception ability of the vision system in the marine environment and lay the foundation for the subsequent target detection of underwater robots.