Image Restoration Using Deep Regulated Convolutional Networks

TL;DR

The proposed Deep Regulated Convolutional Network (RC-Net), a deep network composed of regulated sub-network blocks cascaded by skip-connections, to overcome this bottleneck in designing wider networks and outperform state-of-the-art approaches with significant performance gains in various image restoration tasks while demonstrating promising generalization ability.

Abstract

While the depth of convolutional neural networks has attracted substantial attention in the deep learning research, the width of these networks has recently received greater interest. The width of networks, defined as the size of the receptive fields and the density of the channels, has demonstrated crucial importance in low-level vision tasks such as image denoising and restoration. However, the limited generalization ability, due to the increased width of networks, creates a bottleneck in designing wider networks. In this paper, we propose the Deep Regulated Convolutional Network (RC-Net), a deep network composed of regulated sub-network blocks cascaded by skip-connections, to overcome this bottleneck. Specifically, the Regulated Convolution block (RC-block), featured by a combination of large and small convolution filters, balances the effectiveness of prominent feature extraction and the generalization ability of the network. RC-Nets have several compelling advantages: they embrace diversified features through large-small filter combinations, alleviate the hazy boundary and blurred details in image denoising and super-resolution problems, and stabilize the learning process. Our proposed RC-Nets outperform state-of-the-art approaches with significant performance gains in various image restoration tasks while demonstrating promising generalization ability. The code is available at https://github.com/cswin/RC-Nets.

Figures (12)

• Figure 1: A regulated convolution block with 4 composite units (dotted boxes). The first and third group-squares present $1\times1$ convolution; the second and last ones indicate large and small convolution, respectively. The large convolution is regulated by the small one.
• Figure 2: Comparison of validation error of RC-Nets with different number of blocks, a RC-Net having the $2^{nd}$ layer removed, and WIN during training.
• Figure 3: A deep RC-Net with four RC-blocks. In RC-blocks, the largest, medium and smallest size of circles denote a composite unit using large, small, and $1\times1$ convolution filters, respectively. The $\bigoplus$ represents a summation computing and indicates residual learning.
• Figure 4: Visual results of one image from BSD200-test with $\sigma=10$ along with PSNR(dB) / SSIM.
• Figure 5: Visual results of one image from BSD200-test with $\sigma=70$ along with PSNR(dB) / SSIM.