Face image retrieval across age variation using relevance feedback

Abstract

Given a single face image of a specific person as a query, it is very difficult to retrieve all of his/her images from a personal image collection stored for a long term due to age-related changes in facial appearances. This paper proposes to apply relevance feedback to enhance the performance of image retrieval from the image collections with age variation. Specifically, we propose two types of update schemes: i) query expansion and ii) weight updating and show the effects of each scheme by experiments with two actual image collections. For an image collection, the recall rate improved from 40.8% to 72.5% after five iterations of relevance feedback. © 2010 Springer-Verlag Berlin Heidelberg.