Abstract
We present an end-to-end pipeline for localizing and deskewing one or multiple foreground objects (documents, photos, cards, receipts) placed on a flatbed scanner with a white to near-white background. The method integrates iterative morphological contrast shaping (white / black top-hats), gamma-driven diffusion, statistical region merging (SRM), distance-transform based consolidation, and Voronoi-like label propagation. A unified skew estimation interface produces deskewed, refined crops per object. This document mirrors the structural flow of a companion skew orientation disclosure while remaining self-contained and focused on the flatbed multi-object extension.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.
Recommended Citation
INC, HP, "A Modular Pipeline for Flatbed Content Detection and Localization with Accurate Deskew", Technical Disclosure Commons, (October 23, 2025)
https://www.tdcommons.org/dpubs_series/8778