Image deraining has become a hot topic in the field of computervision. It is the process of removing rain streaks from an image to reconstructa high-quality background. This study aims at improving the performance ofi...Image deraining has become a hot topic in the field of computervision. It is the process of removing rain streaks from an image to reconstructa high-quality background. This study aims at improving the performance ofimage rain streak removal and reducing the disruptive effects caused by rain.To better fit the rain removal task, an innovative image deraining method isproposed, where a kernel prediction network with Unet++ is designed andused to filter rainy images, and rainy-day images are used to estimate thepixel-level kernel for rain removal. To minimize the gap between synthetic andreal data and improve the performance in real rainy image handling, a lossfunction and an effective data optimization method are suggested. In contrastwith other methods, the loss function consists of Structural Similarity Indexloss, edge loss, and L1 loss, and it is adopted to improve performance. Theproposed algorithm can improve the Peak Signal-to-Noise ratio by 1.3% whencompared to conventional approaches. Experimental results indicate that theproposed method can achieve a better efficiency and preserve more imagestructure than several classical methods.展开更多
基金supported by National Natural Science Foundation of China (61772179)Hunan Provincial Natural Science Foundation of China (2022JJ50016,2020JJ4152)+3 种基金the Science and Technology Plan Project of Hunan Province (2016TP1020)Scientific Research Fund of Hunan Provincial Education Department (21B0649)Application-Oriented Characterized Disciplines,Double First-Class University Project of Hunan Province (Xiangjiaotong[2018]469)Discipline Special Research Projects of Hengyang Normal University (Grant No.XKZX21002).
文摘Image deraining has become a hot topic in the field of computervision. It is the process of removing rain streaks from an image to reconstructa high-quality background. This study aims at improving the performance ofimage rain streak removal and reducing the disruptive effects caused by rain.To better fit the rain removal task, an innovative image deraining method isproposed, where a kernel prediction network with Unet++ is designed andused to filter rainy images, and rainy-day images are used to estimate thepixel-level kernel for rain removal. To minimize the gap between synthetic andreal data and improve the performance in real rainy image handling, a lossfunction and an effective data optimization method are suggested. In contrastwith other methods, the loss function consists of Structural Similarity Indexloss, edge loss, and L1 loss, and it is adopted to improve performance. Theproposed algorithm can improve the Peak Signal-to-Noise ratio by 1.3% whencompared to conventional approaches. Experimental results indicate that theproposed method can achieve a better efficiency and preserve more imagestructure than several classical methods.
