Résumé: Single image deraining is an extremely challenging problem since rain streaks severely degrade the image quality and affect the performance of existing outdoor vision tasks. In order to effectively extract the rain streaks features and restore the damaged background texture information, a combining multiscale learning and attention mechanisms densely connected network (CMADNet) is proposed in this paper. Specifically, Multiscale segmentation attention module and DenseNet are employed to construct the overall framework, where the multiscale segmentation attention module aims to use the attention mechanism to learn the feature maps of rain areas, and the DenseNet helps to enhance the feature reuse. Furthermore, considering the important information on multiscale features, a multiscale feature learning module is proposed, where the re-parameterization VGG is used to extract different scale feature maps and effectively characterize the rain streaks features. The experimental results on synthetic data sets and real data sets show that the proposed method is superior to the recent state-of-the-art methods in quantitative index and qualitative analysis.

Langue: Anglais