Better "CMOS" Produces Clearer Images: Learning Space-Variant Blur Estimation for Blind Image Super-Resolution

TL;DR

This paper introduces a new dataset and a novel neural network architecture for blind image super-resolution that effectively handles space-variant blur, improving image quality in real-world scenarios.

Contribution

The paper proposes a new dataset for space-variant blur and a Cross-MOdal fuSion network (CMOS) that estimates blur and semantics simultaneously for better super-resolution.

Findings

Outperforms state-of-the-art methods on new datasets and real-world images.

Achieves higher PSNR and SSIM scores, e.g., +1.91 PSNR on NYUv2-BSR.

Demonstrates the effectiveness of the GIA module for feature interaction.

Abstract

Most of the existing blind image Super-Resolution (SR) methods assume that the blur kernels are space-invariant. However, the blur involved in real applications are usually space-variant due to object motion, out-of-focus, etc., resulting in severe performance drop of the advanced SR methods. To address this problem, we firstly introduce two new datasets with out-of-focus blur, i.e., NYUv2-BSR and Cityscapes-BSR, to support further researches of blind SR with space-variant blur. Based on the datasets, we design a novel Cross-MOdal fuSion network (CMOS) that estimate both blur and semantics simultaneously, which leads to improved SR results. It involves a feature Grouping Interactive Attention (GIA) module to make the two modalities interact more effectively and avoid inconsistency. GIA can also be used for the interaction of other features because of the universality of its structure. Qualitative and quantitative experiments compared with state-of-the-art methods on above datasets and real-world images demonstrate the superiority of our method, e.g., obtaining PSNR/SSIM by +1.91/+0.0048 on NYUv2-BSR than MANet.