A Single Image Dehazing Method Based on Multiple Attention Mechanisms and Multi-Scale Convolution
Shilin Li, Jie Yang · 2024
Haze is a type of atmospheric extinction phenomenon that poses significant challenges to visual systems, such as driving, monitoring, and military reconnaissance. To address residual haze and detail loss in existing dehazing methods, this paper proposes a novel convolutional network model that integrates multiple attention mechanisms and multi-scale feature fusion. Experimental results show the model effectively removes haze and restores color and fine details in areas with substantial haze and rich texture.