Fidelity- and Perception-Aware Local Implicit Attention for Arbitrary-Scale Image Super-Resolution

2026-06-20

AI Summary

This work addresses the limitations of existing methods in arbitrary-scale image super-resolution (ASISR), where conventional approaches often yield overly smooth results and diffusion models tend to generate structural hallucinations. To jointly optimize perceptual quality and reconstruction fidelity, the authors propose the FPLIA framework, which integrates fidelity-guided features into the diffusion process. Central to this approach are the Fidelity-and-Perception Attention Module (FPAM) and the Adaptive Feature Selection Module (FPSM), which synergistically combine local implicit attention, self-attention, and cross-attention mechanisms to enable precise multi-scale feature modulation. Extensive evaluations on standard ASISR benchmarks demonstrate that FPLIA consistently outperforms state-of-the-art methods in both perceptual realism and reconstruction accuracy, with qualitative and quantitative results confirming its effectiveness.

Abstract

Arbitrary-scale image super-resolution (ASISR) aims to reconstruct high-resolution images from low-resolution inputs over a continuous range of upscaling factors. While traditional pixel-regression approaches often produce overly smooth results that lack realistic details, recent diffusion methods can produce sharper and more realistic textures. However, these diffusion techniques frequently introduce the risk of structural hallucinations. To address these issues, we propose Fidelity- and Perception-Aware Local Implicit Attention (FPLIA), a framework that effectively integrates fidelity-oriented features into a diffusion pipeline to produce realistic and faithful reconstructions for ASISR. We introduce a Fidelity and Perception Attention Module (FPAM), which applies both self-attention and cross-attention to fidelity-oriented and perceptual features to enhance representational capacity. To further exploit their complements, we design a Fidelity and Perception Select Module (FPSM) that adaptively selects the most representative features for RGB values prediction. We conduct extensive experiments to validate the effectiveness of these components. Both qualitative and quantitative results show that FPLIA delivers superior perceptual realism while maintaining reconstruction accuracy on standard ASISR benchmarks. The source code is accessible at the following repository: https://github.com/XUSean0118/FPLIA.

Problem

• arbitrary-scale image super-resolution
• fidelity
• perception
• structural hallucinations
• diffusion methods

Innovation

• arbitrary-scale super-resolution
• diffusion models
• fidelity-perception trade-off
• local implicit attention
• feature selection