SVG-to-RDF image semantization

Abstract

The goal of this work is to provide an original (semi-automatic) annotation framework titled SVG-to-RDF whichconverts a collection of raw Scalable vector graphic (SVG) images into a searchable semantic-based RDF graph structure that encodes relevant features and contents. Using a dedicated knowledge base, SVG-to-RDF offers the user possible semantic annotations for each geometric object in the image, based on a combination of shape, color, and position similarity measures. Our method presents several advantages, namely i) achieving complete semantization of image content, ii) allowing semantic-based data search and processing using standard RDF technologies, iii) while being compliant with Web standards (i.e., SVG and RDF) in displaying images and annotation results in any standard Web browser, as well as iv) coping with different application domains. Our solution is of linear complexity in the size of the image and knowledge base structures used. Using our prototype SVG2RDF, several experiments have been conducted on a set of panoramic dental x-ray images to underline our approach’s effectiveness, and its applicability to different application domains.