GFRRN: Explore the Gaps in Single Image Reflection Removal

TL;DR

This paper introduces GFRRN, a novel reflection removal network that employs efficient fine-tuning, label unification, frequency learning, and dynamic attention to improve performance on single image reflection removal tasks.

Contribution

The paper proposes GFRRN, integrating PEFT, label generator, G-AFLB, and DAA to address semantic gaps and label inconsistencies in reflection removal.

Findings

Achieves superior performance over state-of-the-art methods.

Effectively unifies reflection labels for synthetic and real data.

Demonstrates robustness and accuracy in diverse scenarios.

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.