Context-Aware Convolutional Neural Network for Single Image Super-Resolution

Abstract

Super-resolution is an algorithm for reconstructing high-resolution image from low-resolution image. It is one of the more active research topics in the field of computer vision. In recent years, convolutional neural network has made much progress in single image super-resolution. However, the super-resolution algorithms based on convolution network is still difficult to accurately reconstruct the image. In order to improve the reconstruction performance, it is essential to consider the channel information and spatial information of convolution network. We propose a model based on convolutional neural network and attention mechanism, which utilize channel attention modules and spatial attention modules to enhance the flow of information. The convolution network which combines the channel and spatial context information can enhance the network performance. Our model has better ability to utilize context features to more effectively reconstruct the image. Experimental results on several widely used datasets show that our model achieves better reconstruction performance than other model.