Single image rain removal algorithm based on Swin Transformer
Images captured in the rain usually contain many rainmarks,which can obscure the detailed information in the image,making subsequent computer vision tasks such as image classification and object de-tection more difficult.Therefore,it is crucial to remove rain marks from the image and restore the detailed in-formation within them.In response to the above issues,this article proposes a single image rain removal algo-rithm based on Swin Transformer.The algorithm first obtains an approximate rain trace map through the maxi-mum color channel;Then,multi-scale feature extraction and fusion are performed through an encoder decoder structure to obtain better rain removal images.The experimental results show that the structural similarity of the algorithm proposed in this paper on the Rain1200 dataset and Rain1400 dataset is 0.922 and 0.914,re-spectively,and the peak signal-to-noise ratio is 33.28 dB and 31.31 dB,respectively.Compared to the cur-rent mainstream single image rain removal algorithms,the algorithm proposed in this paper performs better in removing rain marks and restoring background details.
single image rain removalSwin Transformerself-attention
万方数据
维普
