CISRDCNN: Super-resolution of compressed images using deep convolutional neural networks

In recent years, many studies have been conducted on image super-resolution (SR). However, to the best of our knowledge, few SR methods are concerned with compressed images. The SR of compressed images is a challenging task due to the complicated compression artifacts that many images suffer from in practice. The intuitive solution for this difficult task is to decouple it into two sequential but independent subproblems, the compression artifacts reduction (CAR) and the SR. Nevertheless, some useful details may be removed in the CAR stage, which is contrary to the goal of SR and makes the SR stage more challenging. In this paper, an end-to-end trainable deep convolutional neural network is designed to perform SR on compressed images, which jointly reduces compression artifacts and improves image resolution. The designed network is named CISRDCNN. Experiments on JPEG images (we take the JPEG as an example in this paper) demonstrate that the proposed CISRDCNN yields state-of-the-art SR performance on commonly used test images and imagesets. The results of CISRDCNN on real low-quality web images are also very impressive with obvious quality improvements. Further, we explore the application of the proposed SR method in low bit-rate image coding, leading to better rate-distortion performance than JPEG.