Super-Resolution using Convolutional Neural Networks without Any Checkerboard Artifacts

TL;DR

This paper proposes a new CNN-based super-resolution method that effectively prevents checkerboard artifacts, a common issue in existing CNN SR techniques, by extending conditions from linear systems to nonlinear CNN architectures.

Contribution

It introduces a novel structure for CNN-based super-resolution that avoids checkerboard artifacts, applicable under multiple loss functions, enhancing image quality without sacrificing performance.

Findings

Successfully avoids checkerboard artifacts in CNN SR methods

Works under mean square error and perceptual loss conditions

Maintains high super-resolution quality

Abstract

It is well-known that a number of excellent super-resolution (SR) methods using convolutional neural networks (CNNs) generate checkerboard artifacts. A condition to avoid the checkerboard artifacts is proposed in this paper. So far, checkerboard artifacts have been mainly studied for linear multirate systems, but the condition to avoid checkerboard artifacts can not be applied to CNNs due to the non-linearity of CNNs. We extend the avoiding condition for CNNs, and apply the proposed structure to some typical SR methods to confirm the effectiveness of the new scheme. Experiment results demonstrate that the proposed structure can perfectly avoid to generate checkerboard artifacts under two loss conditions: mean square error and perceptual loss, while keeping excellent properties that the SR methods have.