Layout Based Information Retrieval from Document Images

This research is intended to develop a layout based retrieval system for document image databases consisting of three phases: 1. At first, intelligent layout analysis algorithm has been designed to extract the layouts the document images physically with their edges and rectangles. 2. Every physically identified layout has been converted into a tree intermediary representation for indexing and storage in layout databases. Later when a query image has been supplied, it retrieves the similar layout images from the layout databases. 3. Finally, a logical layout analysis scheme has been proposed to identify the meaning of the layouts involved in the document images. In intelligent layout analysis system, White Space Analysis technique has been proposed to grab all the white spaces over the image in a single scan over the image with minimum pixel visits, and the white spaces are merged together without the assumptions of heuristics and threshold to segment the layouts. Moreover, two statistical properties have also been proposed in this thesis, to separate the text blocks and images from the identified layouts. In Layout based retrieval system, Tree Representation for physical layouts in a document image and indexing has been proposed in this thesis to enable the retrieval of document images based on layout similarities. This allows a user to find pages with a layout of similar page to the query image supplied by the user based on the tree structure. In logical layout analysis, this thesis attempts to analyze the layouts of the document images logically as title blocks, sub titles, text, images, content lines etc.. to produce more meaningful information from the document images.