Abstract:Blind image deblurring (BID) has been extensively studied in computer vision and adjacent fields. Modern methods for BID can be grouped into two categories: single-instance methods that deal with individual instances using statistical inference and numerical optimization, and data-driven methods that train deep-learning models to deblur future instances directly. Data-driven methods can be free from the difficulty in deriving accurate blur models, but are fundamentally limited by the diversity and quality of the training data -- collecting sufficiently expressive and realistic training data is a standing challenge. In this paper, we focus on single-instance methods that remain competitive and indispensable. However, most such methods do not prescribe how to deal with unknown kernel size and substantial noise, precluding practical deployment. Indeed, we show that several state-of-the-art (SOTA) single-instance methods are unstable when the kernel size is overspecified, and/or the noise level is high. On the positive side, we propose a practical BID method that is stable against both, the first of its kind. Our method builds on the recent ideas of solving inverse problems by integrating the physical models and structured deep neural networks, without extra training data. We introduce several crucial modifications to achieve the desired stability. Extensive empirical tests on standard synthetic datasets, as well as real-world NTIRE2020 and RealBlur datasets, show the superior effectiveness and practicality of our BID method compared to SOTA single-instance as well as data-driven methods. The code of our method is available at: \url{<a class="link-external link-https" href="https://github.com/sun-umn/Blind-Image-Deblurring" rel="external noopener nofollow">this https URL</a>}.

Blind Image Sharpness Assessment And Enhancement Via Deep Auxiliary Learning

Q-DNN: A Quality-Aware Deep Neural Network for Blind Assessment of Enhanced Images.

CS-Net: Deep Multibranch Network Considering Scene Features for Sharpness Assessment of Remote Sensing Images

A shallow convolutional neural network for blind image sharpness assessment

Deep Superpixel-Based Network For Blind Image Quality Assessment

Blind Image Quality Assessment Using a Deep Bilinear Convolutional Neural Network

Blind Surveillance Image Quality Assessment Via Deep Neural Network Combined with the Visual Saliency

A New Blind Image Quality Assessment Based on Pairwise.

Blind Image Quality Assessment Using Multi-Stream Architecture with Spatial and Channel Attention

Blind Image Quality Assessment With Active Inference

AM-BQA: Enhancing blind image quality assessment using attention retractable features and multi-dimensional learning

Blind Image Quality Assessment Based on Multichannel Feature Fusion and Label Transfer.

Opinion-Unaware Blind Image Quality Assessment Using Multi-Scale Deep Feature Statistics

Deep Activation Pooling for Blind Image Quality Assessment

Blind Image Deblurring with Unknown Kernel Size and Substantial Noise

Learning To Blindly Assess Image Quality In The Laboratory And Wild

A blind image super-resolution network guided by kernel estimation and structural prior knowledge

Blind Image Quality Assessment Via Vision-Language Correspondence: A Multitask Learning Perspective.

Deep Image Quality Assessment Driven Single Image Deblurring

End-to-End Blind Image Quality Assessment Using Deep Neural Networks.

Blind Quality Assessment for In-the-wild Images Via Hierarchical Feature Fusion Strategy