Perceptual Quality Optimization of Image Super-Resolution

TL;DR

This paper introduces Efficient-PBAN, a novel deep learning framework that optimizes image super-resolution for human perceptual quality by learning a perceptual quality metric aligned with subjective judgments and integrating it into the training process.

Contribution

The paper presents a new perceptual quality metric and an efficient network that jointly predicts and optimizes for human-perceived image quality in super-resolution tasks.

Findings

Efficient-PBAN correlates strongly with human subjective judgments.

The integrated perceptual loss improves the perceptual quality of super-resolved images.

The approach outperforms existing methods in perceptual quality metrics.

Abstract

Single-image super-resolution (SR) has achieved remarkable progress with deep learning, yet most approaches rely on distortion-oriented losses or heuristic perceptual priors, which often lead to a trade-off between fidelity and visual quality. To address this issue, we propose an \textit{Efficient Perceptual Bi-directional Attention Network (Efficient-PBAN)} that explicitly optimizes SR towards human-preferred quality. Unlike patch-based quality models, Efficient-PBAN avoids extensive patch sampling and enables efficient image-level perception. The proposed framework is trained on our self-constructed SR quality dataset that covers a wide range of state-of-the-art SR methods with corresponding human opinion scores. Using this dataset, Efficient-PBAN learns to predict perceptual quality in a way that correlates strongly with subjective judgments. The learned metric is further integrated into SR training as a differentiable perceptual loss, enabling closed-loop alignment between reconstruction and perceptual assessment. Extensive experiments demonstrate that our approach delivers superior perceptual quality. Code is publicly available at https://github.com/Lighting-YXLI/Efficient-PBAN.