Dehazing algorithms based on credibility fusion
Ludi Wang, Zhaolei Wang, Qinghai Gong · 2021
The mainstream of existing dehazing algorithms can be arranged into two categories: algorithms based on restoration and algorithms based on image enhancement. But the advantages and disadvantages of the two categories are converse. Combining and compensating their characteristics, two fusion-based dehazing methods are designed in this paper, namely Weber Contrast Weighting Method and Credibility-Fusion Adversarial Network Method, with which the dehazing visual effect, details and recognizabilities of dehazed images can all be enhanced. At the same time, the thoroughness of dehazing is also maintained. Compared with the rigid combination in existing papers, this paper fuses softly at each pixel. Experiments suggest that both algorithms proposed in this paper achieve good results and outperform the original algorithms.