Keyword spotting based information Retrieval system for tamil Document images

Abstract

Information Retrieval from Document images is becoming more newlinepopular in todays world due to the construction of paperless offices and newlinedigital libraries using scanned image documents rather than text The existing newlinetools available for retrieving information from text documents could not be newlineapplied to imaged documents directly and therefore the need arose to analyze newlineand study the strategies of information retrieval from document images newlineIn the past decades research on information retrieval has been newlineperformed by concentrating more on text understanding rather than on a newlinecomplete information retrieval system Therefore this thesis aims at newlinedeveloping a complete information retrieval system for Tamil document newlineimages by encompassing major factors such as Layout Analysis Script newlineRecognition Word Image Understanding and Information Retrieval newlineFramework to address the needs of the user in an efficient way newlineInitially Information retrieval has been concentrated on promoting newlinea hybrid approach based physical layout analysis technique for document newlineimages The efficiency of physical layout analysis has been improved in this newlinethesis by proposing a Rectangular White Space Analysis RWSA technique newlinewhich takes white spaces as delimiters to identify heterogeneous layouts both newlineManhattan and Non Manhattan with reduced pixel visits newline newline