RestorMamba: An Enhanced Synergistic State Space Model for Image Restoration
Zeyu Wang, Chen Li, Huiying Xu, Xinzhong Zhu, Xiao Huang, Hongbo Li
Abstract
In this paper, we introduce an image inpainting method based on the State Space Model (SSM), named Restoration Mamba (RestorMamba). This approach incorporates effi-cient long-range dependency modeling within the network, which is particularly suited for the complexities of high-texture and high-resolution image restoration scenarios. To benefit from a broader context while maintaining global receptive fields, we have designed two pivotal modules: Skip Scan and Enhanced Synergistic Mamba (ESM) Block. Our experimental results demonstrate that RestorMamba achieves state-of-the-art performance in tasks such as image deraining and image denoising, encompassing Gaussian grayscale / color denoising and real image denoising.
BibTeX
@inproceedings{icassp2025_restormambaanenh,
title = {RestorMamba: An Enhanced Synergistic State Space Model for Image Restoration},
author = {Zeyu Wang and Chen Li and Huiying Xu and Xinzhong Zhu and Xiao Huang and Hongbo Li},
booktitle = {ICASSP 2025},
year = {2025}
}