Incorporating Uncertainty-Guided and Top-k Codebook Matching for Real-World Blind Image Super-Resolution

2025-06-09

AI Summary

Addressing two key challenges in real-world blind image super-resolution—imprecise codebook matching and texture reconstruction distortion—this paper proposes an uncertainty-guided codebook enhancement framework. First, uncertainty modeling is introduced to identify texture-sensitive regions, enabling adaptive spatial focusing. Second, a Top-k codebook retrieval mechanism is designed to aggregate features from multiple nearest-neighbor codewords, thereby improving matching robustness under unknown degradations. Third, an Align-Attention module is developed to explicitly align low- and high-resolution feature spaces through cross-scale attention with geometric consistency. Evaluated on multiple real-world degradation datasets, the method significantly enhances texture realism and structural fidelity. It achieves superior performance over existing codebook-based approaches in both reconstruction metrics (PSNR/SSIM) and perceptual quality metrics (e.g., LPIPS), establishing a new interpretable and high-fidelity paradigm for codebook utilization in blind super-resolution.

Technology Category

• Computer Vision: Image and Video Retrieval
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Recent advancements in codebook-based real image super-resolution (SR) have shown promising results in real-world applications. The core idea involves matching high-quality image features from a codebook based on low-resolution (LR) image features. However, existing methods face two major challenges: inaccurate feature matching with the codebook and poor texture detail reconstruction. To address these issues, we propose a novel Uncertainty-Guided and Top-k Codebook Matching SR (UGTSR) framework, which incorporates three key components: (1) an uncertainty learning mechanism that guides the model to focus on texture-rich regions, (2) a Top-k feature matching strategy that enhances feature matching accuracy by fusing multiple candidate features, and (3) an Align-Attention module that enhances the alignment of information between LR and HR features. Experimental results demonstrate significant improvements in texture realism and reconstruction fidelity compared to existing methods. We will release the code upon formal publication.

Problem

• Inaccurate feature matching with codebook in image SR
• Poor texture detail reconstruction in real-world SR
• Need for better alignment of LR and HR features

Innovation

• Uncertainty learning for texture-rich regions
• Top-k feature matching for accuracy
• Align-Attention module for feature alignment