The method of component-based image retrieval in document

초록

A document or paper contains various types of images. This paper presents a new method for comparing images in documents or papers. This method consists of three main steps. At the first step, Images are extracted by converting a document into html format. In the second step, the extracted image is divided into subcomponents and the category of each component is labeled. In the final step, descriptive feature values are computed and such values are inserted into a database in the form of descriptors. Experimental results show that the proposed approach works very well in terms of retrieving similar images. It can even retrieve images which are partially similar to a query image.