Image Super-Resolution Using Dense Skip Connections

Recent studies have shown that the performance of single-image super-resolution methods can be significantly boosted by using deep convolutional neural networks. In this study, we present a novel single-image super-resolution method by introducing dense skip connections in a very deep network. In the proposed network, the feature maps of each layer are propagated into all subsequent layers, providing an effective way to combine the low-level features and high-level features to boost the reconstruction performance. In addition, the dense skip connections in the network enable short paths to be built directly from the output to each layer, alleviating the vanishing-gradient problem of very deep networks. Moreover, deconvolution layers are integrated into the network to learn the upsampling filters and to speedup the reconstruction process. Further, the proposed method substantially reduces the number of parameters, enhancing the computational efficiency. We evaluate the proposed method using images from four benchmark datasets and set a new state of the art.

1. Introduction

The recovery of a high resolution (HR) image from a low resolution (LR) version is a highly ill-posed problem since the mapping from LR to HR space can have multiple solutions. When the upscaling factor is large, it becomes very challenging to recover the high-frequency details in image super-resolution (SR). Many SR techniques assume that the high-frequency information is redundant and can be accurately predicted from the low-frequency data. Therefore, it is important to collect useful contextual information in large regions from LR images so that sufficient knowledge can be captured for recovering the high-frequency details in HR images.