Image super-resolution via channel attention and spatial graph convolutional network
作者：
Highlights：
• A novel channel attention and spatial graph convolutional network (CASGCN) for accurate image SR task is proposed.
• An effective channel attention and spatial graph (CASG) block is developed in CASGCN to extract informative features and self-similar information across the channel and spatial dimensions.
• An adjacency matrix in the graph convolutional layer of CASG is generated dynamically by Gram matrix which requires no additional parameters to learn global self-similar patterns.
摘要
•A novel channel attention and spatial graph convolutional network (CASGCN) for accurate image SR task is proposed.•An effective channel attention and spatial graph (CASG) block is developed in CASGCN to extract informative features and self-similar information across the channel and spatial dimensions.•An adjacency matrix in the graph convolutional layer of CASG is generated dynamically by Gram matrix which requires no additional parameters to learn global self-similar patterns.
论文关键词：Image super-resolution,Graph convolutional,Adjacent matrix
论文评审过程：Received 8 April 2020, Revised 19 July 2020, Accepted 14 December 2020, Available online 28 December 2020, Version of Record 28 December 2020.
论文官网地址：https://doi.org/10.1016/j.patcog.2020.107798
