Less is More: Learning Reference Knowledge Using No-Reference Image Quality Assessment

TL;DR

This paper introduces a novel no-reference image quality assessment framework that learns reference knowledge through feature distillation and regularization, achieving superior performance with less input.

Contribution

It proposes a new framework for NR-IQA that learns from non-aligned references using feature distillation and inductive bias regularization, improving accuracy and efficiency.

Findings

Achieves state-of-the-art performance on eight NR-IQA datasets.

Outperforms existing methods with higher PLCC scores.

Utilizes less input data while improving quality assessment accuracy.

Abstract

Image Quality Assessment (IQA) with reference images have achieved great success by imitating the human vision system, in which the image quality is effectively assessed by comparing the query image with its pristine reference image. However, for the images in the wild, it is quite difficult to access accurate reference images. We argue that it is possible to learn reference knowledge under the No-Reference Image Quality Assessment (NR-IQA) setting, which is effective and efficient empirically. Concretely, by innovatively introducing a novel feature distillation method in IQA, we propose a new framework to learn comparative knowledge from non-aligned reference images. And then, to achieve fast convergence and avoid overfitting, we further propose an inductive bias regularization. Such a framework not only solves the congenital defects of NR-IQA but also improves the feature extraction framework, enabling it to express more abundant quality information. Surprisingly, our method utilizes less input while obtaining a more significant improvement compared to the teacher models. Extensive experiments on eight standard NR-IQA datasets demonstrate the superior performance to the state-of-the-art NR-IQA methods, i.e., achieving the PLCC values of 0.917 (vs. 0.884 in LIVEC) and 0.686 (vs. 0.661 in LIVEFB).