Text Block Segmentation for Image Search System

Abstract

This paper presents the text block segmentation for image search system.Web pages are segmented into blocks for getting the text with images. Text block with images are later used as captions of images and hence plays an important factor for image search. Text block segmentation is performed for image indexing using combination of document object model (DOM) Tree and semantic relevancebetween text blocks.Relatedness between text blocks are computed based semantic relevance between terms. DOM tree is widely used in XML parsing. In this paper, HTML tags are used in building DOM tree for web page segmentation. Semantic similarity is computed based on their word hierarchy to the root and hence improving performance of the system. Semantic based text block segmentation is calculated using WordNet dictionary.