Deep learning method for rain streaks removal from single image. Issue 13 (28th May 2020)

• Record Type: Journal Article
• Authors: Wang, Meihua; Chen, Lunbao; Liang, Yun; Huang, Han; Cai, Ruichu
• Is Part Of: Journal of engineering. Volume 2020:Issue 13(2020)
• Page Start: 555
• Page End: 560
• Publication Date: 2020-05-28
• Subjects: learning (artificial intelligence) -- neural nets -- convolution -- image restoration -- image representation -- feature extraction; deep learning method -- rain -- single image -- input image -- convolutional layers -- ROIC layers -- Reusing Original Image Convolutional -- regular convolutional layer -- synthetic images -- real‐world images; Engineering -- Periodicals; Engineering; Electronic journals; Periodicals; 620.005
• DOI: 10.1049/joe.2019.1197
• Languages: English
• ISSNs: 2051-3305

Abstract

Abstract : In this study, a deep learning method for rain streaks removal from a single image is proposed. The main idea is to reuse the original image in the network because the original image can provide more details of the background. These details can be useful for the restoration of the image after rain streaks removal. The authors concatenate the input image of the network and the feature maps generated by the former layers before entering the subsequent layers. The convolutional layers reusing the original image are called ROIC (Reusing Original Image Convolutional) layers. They take the original image as a part of their inputs. The network consists of five convolutional layers: one regular convolutional layer and four ROIC layers. Despite the fact that the network is trained on the synthetic data, experimental results show that the proposed method has comparable performance on both synthetic images and real‐world images to the state‐of‐the‐art methods.