Guiding Perception-Reasoning Closer to Human in Blind Image Quality Assessment

TL;DR

This paper proposes a reinforcement learning approach guided by human annotations to develop a blind image quality assessment model that mimics human perception and reasoning, achieving comparable accuracy and improved interpretability.

Contribution

It introduces a human-guided reinforcement learning framework that enhances both the accuracy and interpretability of BIQA models by incorporating human perception-reasoning data.

Findings

Achieves state-of-the-art correlation metrics in BIQA.

Improves alignment with human reasoning chains as measured by ROUGE-1.

Demonstrates the model's ability to generate human-like explanations.

Abstract

Humans assess image quality through a perception-reasoning cascade, integrating sensory cues with implicit reasoning to form self-consistent judgments. In this work, we investigate how a model can acquire both human-like and self-consistent reasoning capability for blind image quality assessment (BIQA). We first collect human evaluation data that capture several aspects of human perception-reasoning pipeline. Then, we adopt reinforcement learning, using human annotations as reward signals to guide the model toward human-like perception and reasoning. To enable the model to internalize self-consistent reasoning capability, we design a reward that drives the model to infer the image quality purely from self-generated descriptions. Empirically, our approach achieves score prediction performance comparable to state-of-the-art BIQA systems under general metrics, including Pearson and Spearman correlation coefficients. In addition to the rating score, we assess human-model alignment using ROUGE-1 to measure the similarity between model-generated and human perception-reasoning chains. On over 1,000 human-annotated samples, our model reaches a ROUGE-1 score of 0.512 (cf. 0.443 for baseline), indicating substantial coverage of human explanations and marking a step toward human-like interpretable reasoning in BIQA.