Lightweight Implementation of a Transformer-based Single Image Dehaze Network
Jikang Mo, Zongbo Hao, Zixu Tao, Hanqing Lin, Can Yu, Di Lin · 2024
With the rapid advancement of information technology, image dehazing technology has become increasingly important in many fields. In recent years, a multitude of new image dehazing methods have been proposed and have been widely applied across various sectors. Currently, as the Transformer architecture grows in popularity, this paper considers integrating the Transformer network structure into the domain of image dehazing. The paper aims to optimize the traditional U-shaped dehazing network that utilizes semantic segmentation by reducing the number of network layers to achieve a lightweight design while also improving the network's dehazing performance. Ultimately, the performance and effectiveness of the proposed network in this paper have been evaluated on typical datasets.