Image defogging method based on multi-shard hierarchical network

In the field of computer vision, traditional image dehazing methods are difficult to balance detail restoration and colour reproduction. Therefore, this study proposes a hybrid defogging method based on deep multi-shard layered networks. This method innovatively combines multi-shard with layered network technology, utilizing gated convolution technology to achieve fine-grained processing of image data. Finally, the depth image prior algorithm, edge connection algorithm and partial convolution algorithm were used as comparative algorithms for experimental verification. The results showed that the proposed method learned a similarity of 0.024 for the perceived image blocks, while the similarities of the compared algorithms were 0.148, 0.058, and 0.041, respectively. In empirical analysis, the proposed method has an average information entropy of 7.594 on five types of fog images. The above results indicate that this study provides an efficient and high-quality solution for image dehazing.