Efficient and Accurate Multi-Scale Topological Network for Single Image Dehazing

Single image dehazing is a challenging ill-posed problem that has drawn significant attention in the last few years. Recently, convolutional neural networks have achieved great success dehazing. However, it still difficult for these increasingly complex models to recover accurate details from hazy image. In this paper, we pay feature extraction and utilization of input itself. To achieve this, propose Multi-scale Topological Network (MSTN) fully explore features at different scales. Meanwhile, design Feature Fusion Module (MFFM) an Adaptive Selection (AFSM) selection fusion scales, so as progressive This topological network provides large number search paths enable extract abundant well strong fault tolerance robustness. addition, ASFM MFFM can adaptively select important ignore interference information when fusing scale representations. Extensive experiments are conducted demonstrate superiority our method compared with state-of-the-art methods.