DEFNet: Multitasks-based Deep Evidential Fusion Network for Blind Image Quality Assessment

TL;DR

A multitasks-based Deep Evidential Fusion Network (DEFNet) for BIQA, which performs multitask optimization with the assistance of scene and distortion type classification tasks and a novel trustworthy information fusion strategy to achieve a more robust and reliable representation.

Abstract

Blind image quality assessment (BIQA) methods often incorporate auxiliary tasks to improve performance. However, existing approaches face limitations due to insufficient integration and a lack of flexible uncertainty estimation, leading to suboptimal performance. To address these challenges, we propose a multitasks-based Deep Evidential Fusion Network (DEFNet) for BIQA, which performs multitask optimization with the assistance of scene and distortion type classification tasks. To achieve a more robust and reliable representation, we design a novel trustworthy information fusion strategy. It first combines diverse features and patterns across sub-regions to enhance information richness, and then performs local-global information fusion by balancing fine-grained details with coarse-grained context. Moreover, DEFNet exploits advanced uncertainty estimation technique inspired by evidential learning with the help of normal-inverse gamma distribution mixture. Extensive experiments on both synthetic and authentic distortion datasets demonstrate the effectiveness and robustness of the proposed framework. Additional evaluation and analysis are carried out to highlight its strong generalization capability and adaptability to previously unseen scenarios.

Figures (11)

• Figure 1: Comparison between the proposed DEFNet and state-of-the-art methods that utilize auxiliary tasks to assist in BIQA. We propose to include evidential fusion for each task for higher performance and lower uncertainty.
• Figure 2: Overview of the proposed DEFNet framework.
• Figure 3: Overview of the cross sub-region information fusion (top) and local-global information fusion (bottom).
• Figure 4: gMAD competition between UNIQUE and DEFNet in WED. (a) Fixed UNIQUE at low-quality level. (b) Fixed UNIQUE at high-quality level. (c) Fixed DEFNet at low-quality level. (d) Fixed DEFNet at high-quality level.
• Figure 5: Kernel density estimation plots.