MSSNet: Multi-Scale-Stage Network for Single Image Deblurring

Kiyeon Kim      Seungyong Lee     Sunghyun Cho

ECCV 2022 Workshop (AIM 2022)


Most traditional single image deblurring methods before deep learning adopt a coarse-to-fine scheme that estimates a sharp image at a coarse scale and progressively refines it at finer scales. While this scheme has also been adopted in several deep learning-based approaches, recently a number of single-scale approaches have been introduced showing superior performance to previous coarse-to-fine approaches in terms of quality and computation time. In this paper, we revisit the coarse-to-fine scheme and analyze the defects of previous coarse-to-fine approaches. Based on the analysis, we propose Multi-Scale-Stage Network (MSSNet), a novel deep learning-based approach to single image deblurring with our remedies to the defects. MSSNet adopts three remedies: stage configuration reflecting blur scales, an inter-scale information propagation scheme, and a pixel-shuffle-based multi-scale scheme. Our experiments show that our remedies can effectively resolve the defects of previous coarse-to-fine approaches and improve the deblurring performance.


author = {Kim, Kiyeon and Lee, Seungyong and Cho, Sunghyun},
title = {MSSNet: Multi-Scale-Stage Network for Single Image Deblurring},
booktitle = {Proc. of ECCVW (AIM)}
year = {2022},