MAT: Multi-Range Attention Transformer for Efficient Image Super-Resolution

TL;DR

The paper introduces MAT, a novel transformer-based approach for image super-resolution that efficiently captures multi-range features using dilation and sparse attention, outperforming existing models in both accuracy and speed.

Contribution

The paper proposes the Multi-Range Attention Transformer (MAT) with a new MSConvStar module, enabling efficient multi-range feature capture and surpassing state-of-the-art SR models in performance and speed.

Findings

MAT achieves superior super-resolution quality compared to existing models.

MAT is approximately 3.3 times faster than SRFormer-light.

The combination of multi-range attention and MSConvStar enhances feature diversity.

Abstract

Image super-resolution (SR) has significantly advanced through the adoption of Transformer architectures. However, conventional techniques aimed at enlarging the self-attention window to capture broader contexts come with inherent drawbacks, especially the significantly increased computational demands. Moreover, the feature perception within a fixed-size window of existing models restricts the effective receptive field (ERF) and the intermediate feature diversity. We demonstrate that a flexible integration of attention across diverse spatial extents can yield significant performance enhancements. In line with this insight, we introduce Multi-Range Attention Transformer (MAT) for SR tasks. MAT leverages the computational advantages inherent in dilation operation, in conjunction with self-attention mechanism, to facilitate both multi-range attention (MA) and sparse multi-range attention (SMA), enabling efficient capture of both regional and sparse global features. Combined with local feature extraction, MAT adeptly capture dependencies across various spatial ranges, improving the diversity and efficacy of its feature representations. We also introduce the MSConvStar module, which augments the model's ability for multi-range representation learning. Comprehensive experiments show that our MAT exhibits superior performance to existing state-of-the-art SR models with remarkable efficiency (~3.3 faster than SRFormer-light).