Towards efficient image retrieval based on multiple features

WWW image retrieval has been a very challenging research topic in the last decade. Interestingly, WWW images are typically described by both high-level (text) and low-level (visual) high-dimensional features. While most research effort has focused on effectiveness issues, the problems related to efficient retrieval from WWW image databases using an approach based on multiple features have not been seriously addressed. In this paper, we discuss the difficulties in indexing multi-feature image databases, and outline, various indexing approaches for such databases, as well as consider their respective strengths and weaknesses. Corresponding query processing algorithms are also briefly discussed.