Poorly Structured Handwritten Documents Segmentation using Continuous Probabilistic Feature Grammars
Thierry Artières · 2003
This work deals with poorly structured handwritten documents segmentation such as pages of handwritten notes produced with pen-based interfaces. We propose to use a formalism, based on Probabilistic Feature Grammars, that exhibit some interesting features. It allows handling ambiguities and to taking into account contextual information such as spatial relations between objects in the page.