Densely convolutional attention network for image super-resolution

Recently, deep convolutional neural networks (CNNs) with skip-connections in single image super-resolution (SISR) have received great success. However, the difficulty of reconstructing high-resolution images is particularly pronounced for the limited information extracted from input low-resolution images, for which the performance bottleneck of these SR networks still exists. In this paper, we present a novel Densely Convolutional Attention Network (DCAN) for SISR to improve the quality of reconstructed images. Specifically, we propose a convolutional attention mechanism which adds an weighted skip-connection joint the original convolution. Our network can adaptively recalibrate convolutional-wise features by considering interdependencies among convolutional layers. Besides, the proposed network leverage several dense block to enhance the relationship between different convolutional layers, more information could be used sufficiently by this dense block learning method, in order that reconstructed images can obtain richer details and clearer edges. Extensive experiments show that the proposed DCAN could obtain more outstanding performance quantitatively and qualitatively against the state-of-the-art methods while cost less execution time. (C) 2019 Elsevier B.V. All rights reserved.