A Bi-objective optimization model for interactive face retrieval

Abstract

In this paper, based on Bayesian relevance feedback methods, we propose a novel interactive face retrieving model based on two objective functions, one is the Maximum a Posterior (MAP) and the other is maximization of mutual information. The proposed bi-objective optimization model aims at minimizing both the number of interactive iterations and the average length of iterations. Moreover, we deduce a top-bottom search algorithm to solve the proposed. Experiments with real testers prove that the proposed algorithm could largely improve the interactive searching efficiency in face databases. © 2011 Springer-Verlag Berlin Heidelberg.