MatIR: A Hybrid Mamba-Transformer Image Restoration Model

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wen, Juan, Hou, Weiyan, Van Gool, Luc, Timofte, Radu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915141964005376
author Wen, Juan
Hou, Weiyan
Van Gool, Luc
Timofte, Radu
author_facet Wen, Juan
Hou, Weiyan
Van Gool, Luc
Timofte, Radu
contents In recent years, Transformers-based models have made significant progress in the field of image restoration by leveraging their inherent ability to capture complex contextual features. Recently, Mamba models have made a splash in the field of computer vision due to their ability to handle long-range dependencies and their significant computational efficiency compared to Transformers. However, Mamba currently lags behind Transformers in contextual learning capabilities. To overcome the limitations of these two models, we propose a Mamba-Transformer hybrid image restoration model called MatIR. Specifically, MatIR cross-cycles the blocks of the Transformer layer and the Mamba layer to extract features, thereby taking full advantage of the advantages of the two architectures. In the Mamba module, we introduce the Image Inpainting State Space (IRSS) module, which traverses along four scan paths to achieve efficient processing of long sequence data. In the Transformer module, we combine triangular window-based local attention with channel-based global attention to effectively activate the attention mechanism over a wider range of image pixels. Extensive experimental results and ablation studies demonstrate the effectiveness of our approach.
format Preprint
id arxiv_https___arxiv_org_abs_2501_18401
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MatIR: A Hybrid Mamba-Transformer Image Restoration Model
Wen, Juan
Hou, Weiyan
Van Gool, Luc
Timofte, Radu
Computer Vision and Pattern Recognition
In recent years, Transformers-based models have made significant progress in the field of image restoration by leveraging their inherent ability to capture complex contextual features. Recently, Mamba models have made a splash in the field of computer vision due to their ability to handle long-range dependencies and their significant computational efficiency compared to Transformers. However, Mamba currently lags behind Transformers in contextual learning capabilities. To overcome the limitations of these two models, we propose a Mamba-Transformer hybrid image restoration model called MatIR. Specifically, MatIR cross-cycles the blocks of the Transformer layer and the Mamba layer to extract features, thereby taking full advantage of the advantages of the two architectures. In the Mamba module, we introduce the Image Inpainting State Space (IRSS) module, which traverses along four scan paths to achieve efficient processing of long sequence data. In the Transformer module, we combine triangular window-based local attention with channel-based global attention to effectively activate the attention mechanism over a wider range of image pixels. Extensive experimental results and ablation studies demonstrate the effectiveness of our approach.
title MatIR: A Hybrid Mamba-Transformer Image Restoration Model
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2501.18401