Learning Enriched Features via Selective State Spaces Model for Efficient Image Deblurring

29 Mar 2024  ·  Hu Gao, Depeng Dang ·

Image deblurring aims to restore a high-quality image from its corresponding blurred. The emergence of CNNs and Transformers has enabled significant progress. However, these methods often face the dilemma between eliminating long-range degradation perturbations and maintaining computational efficiency. While the selective state space model (SSM) shows promise in modeling long-range dependencies with linear complexity, it also encounters challenges such as local pixel forgetting and channel redundancy. To address this issue, we propose an efficient image deblurring network that leverages selective state spaces model to aggregate enriched and accurate features. Specifically, we introduce an aggregate local and global information block (ALGBlock) designed to effectively capture and integrate both local invariant properties and non-local information. The ALGBlock comprises two primary modules: a module for capturing local and global features (CLGF), and a feature aggregation module (FA). The CLGF module is composed of two branches: the global branch captures long-range dependency features via a selective state spaces model, while the local branch employs simplified channel attention to model local connectivity, thereby reducing local pixel forgetting and channel redundancy. In addition, we design a FA module to accentuate the local part by recalibrating the weight during the aggregation of the two branches for restoration. Experimental results demonstrate that the proposed method outperforms state-of-the-art approaches on widely used benchmarks.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Defocus Deblurring DPD ALGNet Combined PSNR 26.45 # 1
Combined SSIM 0.821 # 2
LPIPS 0.186 # 4
Image Deblurring GoPro ALGNet-B PSNR 34.05 # 2
SSIM 0.969 # 1
Image Deblurring HIDE ALGNet-B PSNR 31.68 # 1
Deblurring RealBlur-J ALGNet SSIM (sRGB) 0.946 # 1
PSNR (sRGB) 32.94 # 1
Deblurring RealBlur-J (trained on GoPro) ALGNet PSNR (sRGB) 29.12 # 1
SSIM (sRGB) 0.886 # 1
Deblurring RealBlur-R ALGNet PSNR (sRGB) 41.16 # 1
SSIM (sRGB) 0.981 # 1
Deblurring RealBlur-R (trained on GoPro) ALGNet PSNR (sRGB) 36.35 # 1
SSIM (sRGB) 0.961 # 1

Methods