Two Heads Better than One: Dual Degradation Representation for Blind Super-Resolution

2025-11-21

AI Summary

Blind super-resolution (SR) suffers from performance degradation due to unknown degradations—e.g., blur and noise—whose coupled nature hinders accurate modeling. To address this, we propose a dual-branch degradation extraction network that achieves the first unsupervised disentanglement of blur and noise components. The extracted degradation representation serves as a regularization module, optimized via self-supervision by minimizing the discrepancy between SR residuals and corresponding high-resolution (HR) image differences. Furthermore, we introduce a degradation-aware adaptive feature modulation mechanism to dynamically guide the SR network in detail restoration. Our method requires neither degradation priors nor paired training data, and features lightweight architecture with strong generalization capability. Extensive experiments on multiple real-world and synthetic blind SR benchmarks demonstrate state-of-the-art performance, significantly improving reconstruction quality and robustness under complex, composite degradations.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Unsupervised & Self-Supervised Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Previous methods have demonstrated remarkable performance in single image super-resolution (SISR) tasks with known and fixed degradation (e.g., bicubic downsampling). However, when the actual degradation deviates from these assumptions, these methods may experience significant declines in performance. In this paper, we propose a Dual Branch Degradation Extractor Network to address the blind SR problem. While some blind SR methods assume noise-free degradation and others do not explicitly consider the presence of noise in the degradation model, our approach predicts two unsupervised degradation embeddings that represent blurry and noisy information. The SR network can then be adapted to blur embedding and noise embedding in distinct ways. Furthermore, we treat the degradation extractor as a regularizer to capitalize on differences between SR and HR images. Extensive experiments on several benchmarks demonstrate our method achieves SOTA performance in the blind SR problem.

Problem

• Addresses blind super-resolution with unknown degradation types
• Predicts unsupervised blur and noise embeddings separately
• Uses degradation extractor as regularizer for SR-HR differences

Innovation

• Dual Branch Degradation Extractor Network for blind SR
• Predicts unsupervised blur and noise degradation embeddings
• Treats degradation extractor as regularizer using SR-HR differences