DGIQA: Depth-guided Feature Attention and Refinement for Generalizable Image Quality Assessment

2025-05-29

AI Summary

To address the poor generalization of no-reference image quality assessment (NR-IQA) under unknown natural distortions—such as low illumination, haze, and lens flare—this paper proposes a depth-guided cross-modal attention and refinement framework. Methodologically, it introduces (1) Depth-CAR, the first depth-estimation-guided architecture for scene-aware representation learning; (2) the Transformer-CNN Bridge (TCB), enabling efficient fusion of global semantics and local details; and (3) a multi-scale feature refinement module with attention-weighted projection, balancing accuracy and computational efficiency. The proposed method achieves state-of-the-art performance on both synthetic and real-world distortion datasets. It demonstrates superior cross-dataset generalization capability compared to existing approaches, while maintaining high inference speed and training efficiency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

A long-held challenge in no-reference image quality assessment (NR-IQA) learning from human subjective perception is the lack of objective generalization to unseen natural distortions. To address this, we integrate a novel Depth-Guided cross-attention and refinement (Depth-CAR) mechanism, which distills scene depth and spatial features into a structure-aware representation for improved NR-IQA. This brings in the knowledge of object saliency and relative contrast of the scene for more discriminative feature learning. Additionally, we introduce the idea of TCB (Transformer-CNN Bridge) to fuse high-level global contextual dependencies from a transformer backbone with local spatial features captured by a set of hierarchical CNN (convolutional neural network) layers. We implement TCB and Depth-CAR as multimodal attention-based projection functions to select the most informative features, which also improve training time and inference efficiency. Experimental results demonstrate that our proposed DGIQA model achieves state-of-the-art (SOTA) performance on both synthetic and authentic benchmark datasets. More importantly, DGIQA outperforms SOTA models on cross-dataset evaluations as well as in assessing natural image distortions such as low-light effects, hazy conditions, and lens flares.

Problem

• Improving generalization in no-reference image quality assessment (NR-IQA)
• Enhancing feature learning with depth-guided attention and refinement
• Fusing global and local features for better distortion assessment

Innovation

• Depth-Guided cross-attention and refinement mechanism
• Transformer-CNN Bridge for feature fusion
• Multimodal attention-based projection functions