Stochastic language models for style-directed layout analysis of document images

摘要: Image segmentation is an important component of any document image analysis system. While many algorithms exist in the literature, very few i) allow users to specify physical style, and ii) incorporate user-specified style information into algorithm's objective function that be minimized. We describe a algorithm models document's structure as hierarchical where each node describes region using stochastic regular grammar. The exact form hierarchy language specified by user, while probabilities associated with transitions are estimated from groundtruth data. demonstrate on images bilingual dictionaries.