The Visual Object Information Retrieval (VOIR) system described in this paper implements an image retrieval approach that combines two layers, the conceptual and the visual layer. It uses terms from a textual thesaurus to represent the conceptual information and also works with image regions, the visual information. The terms are related with the image regions through a weighted association enabling the execution of concept-level queries. VOIR uses region-based relevance feedback to improve the quality of the results in each query session and to discover new associations between text and image. This paper describes a user-centred and task-oriented comparative evaluation of VOIR which was undertaken considering three distinct versions of VOIR: a full-fledge version; one supporting relevance feedback only at image level; and a third version not supporting relevance feedback at all. The evaluation performed showed the usefulness of region based relevance feedback in the context of VOIR prototype.
@article{arxiv.0809.4834,
title = {Relevance Feedback in Conceptual Image Retrieval: A User Evaluation},
author = {Jose Torres and Luis Paulo Reis},
journal= {arXiv preprint arXiv:0809.4834},
year = {2008}
}