融合全局和局部特征的单幅图像去雨方法

Single Image Deraining Method Fusing Global and Local Features

针对基于卷积神经网络的去雨方法感受野受限的问题,结合Swin Transformer和卷积神经网络各自的优势,提出了一种融合全局和局部特征的单幅图像去雨方法。首先通过卷积神经网络对图像的局部特征进行初步提取;其次通过基于Swin Transformer的多支路网络对不同特征空间内的全局信息进行学习;最后将提取出的多支路全局特征与局部特征进行融合,实现无雨图像的恢复。在多个数据集上与多种主流单幅图像去雨方法进行了对比实验。结果表明,所提方法生成的结果在峰值信噪比和结构相似性指标上都具有一定优势,验证了所提方法在图像去雨任务上的有效性。

In order to solve the problem of limited receptive field of deraining methods based on convolution neural network,we propose a single image deraining method using swin transformer to fuse global and local features,which combines the advantages of swin transformer and convolution neural network.Firstly,the local features of the image is extracted by convolution neural network.Secondly,the global information in different feature spaces is learned by multi-branch network based on Swin Transformer.Finally,the extracted multi-branch global features and local features are fused to realize the restoration of clean image.Compared with several state-of-the-art single image deraining methods on multiple datasets,the results of the proposed network have competitive performance in terms of peak signal-to-noise ratio and structural similarity index,which verifies the effectiveness of the proposed method in single image deraining task.