Unconstrained Tight Structure Extraction Using Voronoi Tesselation on Document Images.
Panduranga Naidu Nagabhushan, Sahana D. Gowda, R. K. Bharathi · MLMTA · 2006
Document structure is the intermediary result obtained through page segmentation, which is used in the analysis of the document image. The structure serves the purpose of extracting the shape of the document from paragraph up to character level in a hierarchical exploratory methodology for understanding the layout structure of the document image. The extracted layout forms a dominant feature which plays a vital role in categorization, equivalence, ranking and retrieval of documents without reading. The main theme of this paper is to obtain the structure of the document, where entities possess unconst rained layout due to different contents. The Voronoi tessellation on document image with the points of interest being every high pixel in the image helps in recognizing the structure of the entities in the document image. New techniques known as Spring Force is being implemented to grab the point of interest present in the unbounded region and the exterior points which frame the exterior boundary of the entities in the document image to obtain a tight structure. The result of spring force technique is not only obtaining a tight structure for entities, but also to obtain the structure of the blank space in the document image. This method works successfully on all types of docum ent images with Non-Manhattan layouts with arbitrary skew.