XYScanNet: A State Space Model for Single Image Deblurring

TL;DR

XYScanNet introduces a novel state space model with a slice-and-scan strategy and a vision state space module, significantly improving perceptual quality in single image deblurring while maintaining competitive distortion metrics.

Contribution

The paper proposes a new slice-and-scan strategy and a vision state space module, advancing state space models for improved image deblurring performance.

Findings

Enhances KID by 17% over nearest competitor

Maintains competitive distortion metrics

Significantly improves perceptual quality

Abstract

Deep state-space models (SSMs), like recent Mamba architectures, are emerging as a promising alternative to CNN and Transformer networks. Existing Mamba-based restoration methods process visual data by leveraging a flatten-and-scan strategy that converts image patches into a 1D sequence before scanning. However, this scanning paradigm ignores local pixel dependencies and introduces spatial misalignment by positioning distant pixels incorrectly adjacent, which reduces local noise-awareness and degrades image sharpness in low-level vision tasks. To overcome these issues, we propose a novel slice-and-scan strategy that alternates scanning along intra- and inter-slices. We further design a new Vision State Space Module (VSSM) for image deblurring, and tackle the inefficiency challenges of the current Mamba-based vision module. Building upon this, we develop XYScanNet, an SSM architecture integrated with a lightweight feature fusion module for enhanced image deblurring. XYScanNet, maintains competitive distortion metrics and significantly improves perceptual performance. Experimental results show that XYScanNet enhances KID by $17\%$ compared to the nearest competitor.