MatteViT: High-Frequency-Aware Document Shadow Removal with Shadow Matte Guidance

2025-12-09

AI Summary

Document shadow removal is critical for enhancing digital document quality and OCR performance, yet existing methods often degrade high-frequency details such as text edges. To address this, we propose a vision Transformer-based shadow removal framework. Our method introduces a High-Frequency Enhancement Module (HFAM) to strengthen frequency-domain detail representation and pioneers a continuous brightness-driven shadow mask guidance mechanism—enabling, for the first time within a Transformer architecture, synergistic integration of frequency-domain enhancement and spatially precise localization. Coupled with multi-stage feature fusion and a shadow generator trained on a self-constructed mask dataset, our approach significantly improves fine-grained structural recovery. Evaluated on the RDD and Kligler benchmarks, it achieves state-of-the-art performance, with substantial gains in OCR accuracy—demonstrating both robustness and practical applicability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Manipulation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Document shadow removal is essential for enhancing the clarity of digitized documents. Preserving high-frequency details (e.g., text edges and lines) is critical in this process because shadows often obscure or distort fine structures. This paper proposes a matte vision transformer (MatteViT), a novel shadow removal framework that applies spatial and frequency-domain information to eliminate shadows while preserving fine-grained structural details. To effectively retain these details, we employ two preservation strategies. First, our method introduces a lightweight high-frequency amplification module (HFAM) that decomposes and adaptively amplifies high-frequency components. Second, we present a continuous luminance-based shadow matte, generated using a custom-built matte dataset and shadow matte generator, which provides precise spatial guidance from the earliest processing stage. These strategies enable the model to accurately identify fine-grained regions and restore them with high fidelity. Extensive experiments on public benchmarks (RDD and Kligler) demonstrate that MatteViT achieves state-of-the-art performance, providing a robust and practical solution for real-world document shadow removal. Furthermore, the proposed method better preserves text-level details in downstream tasks, such as optical character recognition, improving recognition performance over prior methods.

Problem

• Removes shadows from digitized documents to enhance clarity
• Preserves high-frequency details like text edges and lines
• Improves optical character recognition performance in downstream tasks

Innovation

• MatteViT uses spatial and frequency-domain transformer for shadow removal
• HFAM module amplifies high-frequency components to preserve details
• Shadow matte provides spatial guidance from early processing stage