Technology Category
Computer Vision: Learning & Optimization for CVIntelligent Robots: Localization, Mapping, and NavigationMachine Learning: Calibration & Uncertainty Quantification
BookNet: Book Image Rectification via Cross-Page Attention Network
🤖 AI Summary
This work addresses the challenge of asymmetric geometric distortion commonly observed in book spread images due to binding, which existing single-page rectification methods fail to model due to their inability to capture inter-page coupling. To this end, we propose BookNet, the first end-to-end framework for dual-page book image rectification, featuring a dual-branch architecture with a novel cross-page attention mechanism that jointly estimates deformation flows for both left and right pages as well as the fully unfolded layout, thereby explicitly modeling geometric dependencies between pages. We introduce the first cross-page attention mechanism and construct Book3D, the first large-scale synthetic dataset, along with Book100, a real-world benchmark for evaluation. Experiments demonstrate that BookNet significantly outperforms existing approaches, and both the code and datasets will be publicly released.
Technology Category
Application Category
📝 Abstract
Book image rectification presents unique challenges in document image processing due to complex geometric distortions from binding constraints, where left and right pages exhibit distinctly asymmetric curvature patterns. However, existing single-page document image rectification methods fail to capture the coupled geometric relationships between adjacent pages in books. In this work, we introduce BookNet, the first end-to-end deep learning framework specifically designed for dual-page book image rectification. BookNet adopts a dual-branch architecture with cross-page attention mechanisms, enabling it to estimate warping flows for both individual pages and the complete book spread, explicitly modeling how left and right pages influence each other. Moreover, to address the absence of specialized datasets, we present Book3D, a large-scale synthetic dataset for training, and Book100, a comprehensive real-world benchmark for evaluation. Extensive experiments demonstrate that BookNet outperforms existing state-of-the-art methods on book image rectification. Code and dataset will be made publicly available.
Problem
Research questions and friction points this paper is trying to address.
book image rectification
geometric distortion
cross-page relationship
document image processing
asymmetric curvature
Innovation
Methods, ideas, or system contributions that make the work stand out.
cross-page attention
book image rectification
dual-branch architecture
geometric distortion modeling
synthetic dataset
🔎 Similar Papers
No similar papers found.
