Boosting Image Quality Assessment Performance: Unsupervised Score Fusion by Deep Maximum a Posteriori Estimation

2026-05-28

AI Summary

This work addresses the limited generalization of existing single-image quality assessment (IQA) models across diverse content and distortion types, which stems from inherent design biases. To overcome this limitation, the study introduces deep maximum a posteriori (MAP) estimation into unsupervised IQA score fusion—a first in the field—leveraging fine-grained uncertainty modeling to adaptively weight predictions from multiple IQA models. The proposed framework operates without labeled data, automatically suppressing outputs from less reliable models while amplifying those with high predictive confidence. Experimental results demonstrate that the method significantly outperforms both individual IQA models and alternative fusion strategies across multiple benchmarks, and further exhibits the capability to automatically discard poorly performing models.

Abstract

Over the past decades, numerous Image Quality Assessment (IQA) models have emerged, aiming to predict the perceptual quality of images. However, individual models are often biased toward certain types of image content or distortions, depending on the design principle and process. An intuitive idea is to harness the strengths and mitigate the weaknesses of each IQA model, by fusing the scores of multiple models into a stronger one. Here we make one of the first attempts to seek an optimal solution for the idea and propose a general framework for unsupervised IQA score fusion using deep Maximum a Posteriori (MAP) estimation. The proposed model conducts fine-grained uncertainty estimation at the score level to increase the accuracy and reduce the uncertainty in fused predictions. Comprehensive experiments demonstrate the superiority of the proposed model over individual IQA models and other fusion methods. It also exhibits an interesting capability of rejecting ``bad" models in the fusion process.

Problem

• Image Quality Assessment
• Score Fusion
• Model Bias
• Perceptual Quality
• Unsupervised Learning

Innovation

• unsupervised score fusion
• deep Maximum a Posteriori estimation
• image quality assessment
• uncertainty estimation
• model rejection