Integration of Novel Image Based Features into Markov Random Field Model for Information Retrieval

Abstract

This paper develops a general and formal frame-work for the ranking of web documents by considering the multimedia information contained in these documents. Multimedia information is mostly related to images and videos. Ranking can be treated as a combination of static and dynamic ranking. In this paper, we have described a static ranking method based on the analysis of the images present in the web document and integrated it into the ranking framework which is based on Markov Random Field Model, which combines both static and dynamic ranking. We have described a novel metric DQEV(Document Quality Enhancing Value), based on the images present in a web document and the value of DQEV indicates to what extent the images in a web document increase its value. Integration of an Image Search Engine has also been proposed for the computation of DQEV. It has also been shown that the integration of DQEV can increase the effectiveness of the ranking system.