Publications

Advanced search

Abstract

Mauricio Villegas, Luis A. Leiva, Roberto Paredes. Interactive Image Retrieval Based on Relevance Feedback. Multimodal Interaction in Image and Video Applications. Springer Berlin Heidelberg. 2013. Vol. 48 Intelligent Systems Reference Library, pp. 83-109.

This chapter presents a prototype of a web image search engine that implements four approaches to improve the performance of interactive image retrieval systems. The first approach is classic relevance feedback, which relies on user feedback and image dissimilarities in order to provide better retrievals in an iterative process. It adopts a probabilistic model which leads to maximizing the relevance of the images retrieved. The second approach is based on user relevance feedback as well, but the attention is focused on combining several information sources to the retrieval mechanism. In particular, we propose a retrieval technique that combines both visual and textual features using dynamic late fusion. The third and fourth approaches are query refinement and tag cloud, both consisting of leveraging the information derived from the relevance feedback and the (textual) image annotations. In the former, a refinement of the initial textual query is suggested. In the latter, a tag cloud is given to provide an overall topic formation related to the user's image selection.