CASISR: Circular Arbitrary-Scale Image Super-Resolution

2026-05-04

AI Summary

Existing arbitrary-scale image super-resolution methods suffer from limited generalization under data-scarce training conditions. This work proposes Circular Arbitrary-Scale Image Super-Resolution (CASISR), which introduces, for the first time, a closed-loop control perspective to this domain by constructing a nonlinear recurrent system that integrates a learnable degradation model with a super-resolution module. The method’s theoretical foundation is established through conditional probability modeling, and its stability is analyzed via Taylor series expansion. Additionally, first- and second-order absolute difference images are introduced to evaluate reconstruction fidelity. Extensive experiments demonstrate that CASISR significantly outperforms eight state-of-the-art methods across diverse scaling factors—particularly fractional scales—with especially notable gains on images rich in high-frequency details such as text and stripes.

Abstract

The generalization performance (GP) of deep learning-based arbitrary-scale image super-resolution (ASISR) methods is subject to limited training datasets and unlimited testing datasets. It is vitally significant to enhance the GP of the pretrained ASISR models by making full use of the testing samples. The ASISR models usually employ an open-loop architecture from low-resolution (LR) images to super-resolution (SR) images. The degradation model from SR samples to LR samples is known bicubic down-sampling for the classical ASISR, is supposed down-sampling with additive random noise for the blind ASISR, and is learnable for the real-world ASISR. Combining the ASISR and degradation models, it is potentially possible to adopt a closed-loop architecture based on the automatic control theory for strengthening the GP of the ASISR methods. Therefore, this paper proposes a closed-loop architecture, circular ASISR (CASISR), to lift the capability of image reconstruction. A mathematical nonlinear loop equation is established to describe the CASISR, the reasonability of the CASISR is proven by conditional probability theory, and the stability of the CASISR is proven by Taylor series approximation. The first-order and second-order absolute difference images are defined to compare the image reconstruction performance of the ASISR and the CASISR methods. Comprehensive simulation experiments show that the proposed CASISR approach outperforms the eight state-of-the-art ASISR approaches in the quality of image reconstruction. Especially, the proposed CASISR is extraordinarily suitable for fractional SR scale factors and is extremely effective for text and stripe images with drastically changed edges.

Problem

• arbitrary-scale image super-resolution
• generalization performance
• closed-loop architecture
• fractional scale factors
• edge-rich images

Innovation

• Circular ASISR
• Closed-loop architecture
• Arbitrary-scale super-resolution
• Generalization performance
• Degradation model