Progressive Focused Transformer for Single Image Super-Resolution

2025-03-26

AI Summary

To address computational redundancy and performance bottlenecks in Transformer-based single-image super-resolution (SISR) caused by global self-attention, this paper proposes Progressive Focusing Attention (PFA). PFA dynamically identifies and focuses on salient tokens across multi-scale feature layers by establishing cross-layer attention map correlations, jointly enabling attention sparsification, token importance prediction, and dynamic similarity-based pruning. It is the first work to introduce a layer-wise focusing paradigm into SISR Transformers, preserving modeling capacity while significantly improving efficiency. On multiple standard SISR benchmarks, PFA achieves state-of-the-art (SOTA) performance: PSNR gains of 0.15–0.32 dB over prior methods, with approximately 37% reduction in FLOPs—demonstrating superior accuracy–efficiency trade-offs.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Transformer-based methods have achieved remarkable results in image super-resolution tasks because they can capture non-local dependencies in low-quality input images. However, this feature-intensive modeling approach is computationally expensive because it calculates the similarities between numerous features that are irrelevant to the query features when obtaining attention weights. These unnecessary similarity calculations not only degrade the reconstruction performance but also introduce significant computational overhead. How to accurately identify the features that are important to the current query features and avoid similarity calculations between irrelevant features remains an urgent problem. To address this issue, we propose a novel and effective Progressive Focused Transformer (PFT) that links all isolated attention maps in the network through Progressive Focused Attention (PFA) to focus attention on the most important tokens. PFA not only enables the network to capture more critical similar features, but also significantly reduces the computational cost of the overall network by filtering out irrelevant features before calculating similarities. Extensive experiments demonstrate the effectiveness of the proposed method, achieving state-of-the-art performance on various single image super-resolution benchmarks.

Problem

• Reduce computational cost in Transformer-based super-resolution
• Filter irrelevant features before similarity calculations
• Improve attention on critical tokens for better performance

Innovation

• Progressive Focused Transformer for efficient super-resolution
• Progressive Focused Attention to prioritize important tokens
• Reduces computational cost by filtering irrelevant features