GFRRN: Explore the Gaps in Single Image Reflection Removal

2026-02-26

AI Summary

This work addresses the challenges in single-image reflection removal arising from semantic misalignment between pre-trained models and the target task, as well as inconsistencies in reflection labels between synthetic and real-world data. To bridge these gaps, the authors propose the Gap-Free Reflection Removal Network (GFRRN), which introduces a learnable Mona layer for parameter-efficient fine-tuning to align semantic representations. A label generator is designed to unify the reflection label space across synthetic and real data. Furthermore, GFRRN integrates a Gaussian Adaptive Frequency Learning Block (G-AFLB) and a Dynamic Proxy Attention (DAA) mechanism to adaptively fuse frequency-domain priors and dynamically model both intra- and inter-window dependencies, respectively. Extensive experiments demonstrate that GFRRN significantly outperforms existing methods across multiple benchmarks, achieving superior removal performance and enhanced generalization capability.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Prior dual-stream methods with the feature interaction mechanism have achieved remarkable performance in single image reflection removal (SIRR). However, they often struggle with (1) semantic understanding gap between the features of pre-trained models and those of reflection removal models, and (2) reflection label inconsistencies between synthetic and real-world training data. In this work, we first adopt the parameter efficient fine-tuning (PEFT) strategy by integrating several learnable Mona layers into the pre-trained model to align the training directions. Then, a label generator is designed to unify the reflection labels for both synthetic and real-world data. In addition, a Gaussian-based Adaptive Frequency Learning Block (G-AFLB) is proposed to adaptively learn and fuse the frequency priors, and a Dynamic Agent Attention (DAA) is employed as an alternative to window-based attention by dynamically modeling the significance levels across windows (inter-) and within an individual window (intra-). These components constitute our proposed Gap-Free Reflection Removal Network (GFRRN). Extensive experiments demonstrate the effectiveness of our GFRRN, achieving superior performance against state-of-the-art SIRR methods.

Problem

• single image reflection removal
• semantic understanding gap
• label inconsistency
• synthetic and real-world data

Innovation

• Parameter Efficient Fine-Tuning
• Label Consistency
• Adaptive Frequency Learning
• Dynamic Agent Attention
• Single Image Reflection Removal