DPAFNet:Dual Path Attention Fusion Network for Single Image Deraining
This work addresses image quality degradation in rainy conditions for imaging systems, representing an incremental advancement in low-level vision tasks.
The paper tackles the problem of single image deraining by proposing a dual-branch attention fusion network to address the limitation of single-branch neural networks in multidimensional feature fusion, achieving improved performance as validated through ablation and comparison experiments.
Rainy weather will have a significant impact on the regular operation of the imaging system. Based on this premise, image rain removal has always been a popular branch of low-level visual tasks, especially methods using deep neural networks. However, most neural networks are but-branched, such as only using convolutional neural networks or Transformers, which is unfavourable for the multidimensional fusion of image features. In order to solve this problem, this paper proposes a dual-branch attention fusion network. Firstly, a two-branch network structure is proposed. Secondly, an attention fusion module is proposed to selectively fuse the features extracted by the two branches rather than simply adding them. Finally, complete ablation experiments and sufficient comparison experiments prove the rationality and effectiveness of the proposed method.