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

Abstract

In recent years, much research has been conducted on image super-resolution (SR). To the best of our knowledge, however, few SR methods were concerned with compressed images. The SR of compressed images is a challenging task due to the complicated compression artifacts, while many images suffer from them in practice. The intuitive solution for this difficult task is to decouple it into two sequential but independent subproblems, i.e., compression artifacts reduction (CAR) and SR. Nevertheless, some useful details may be removed in 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 (CISRD-CNN), which reduces compression artifacts and improves image resolution jointly. Experiments on compressed images produced by JPEG (we take the JPEG as an example in this paper) demonstrate that the proposed CISRD-CNN 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 enhancement. Further, we explore the application of the proposed SR method in low bit-rate image coding, leading to better rate-distortion performance than JPEG.

A New Optimized Recurrent Feedback Deep Convolutional Neural Net for Image Super Resolution

“…Freely available SET5, SET14, BCD100 [14], [17] Datasets are used for experimentation. Each dataset is a collection of (LR,HR) image pairs.…”

“…To evaluate the performance of proposed system, it is compared with recent SISR techniques: CISRDCNN [14], VDSR [17] and Bicubic Interpolation. Peak signal to noise ratio (PSNR) and structural similarity index (SSIM) are used as performance evaluators.…”

“…Compressed images can be super resolved & HR image can be reconstructed by using compressed image deep convolutional neural networks technique (CISRDCNN) [14]. It includes Residual Learning method for addressing performance degradation problem, Batch Normalization to fasten the training process & uses ReLU as an activation function for DCNN.…”

“…Image restoration can be performed at both the ends (LR & HR). The prior methods, which addressed the SR problem using deep convolutional neural networks like SRCNN [15], FSRCNN [16], VDSR [17] & DnCNN [18] have not considered the compression artifacts in SR problem, which is addressed in [14]. But image restoration is dependent upon characteristics of input.…”

“…The works [14], [17] proposed novel SR techniques using convolutional neural networks for goal of improving peak signal to noise ratio of output images. The technique [14] was focused on compressed images, while the other technique [17] was focused on increasing network layers to achieve maximum accuracy. But as per our knowledge, very few of these techniques tried to improve structural similarity between ground truth HR and predicted HR.…”