Boosting colored local features for generic object recognition

Abstract

Inclusion of local color information in generic object recognition is ignored by almost all approaches, although it is important and can improve the recognition performance. In this paper, we present a generic object recognition approach using boosting as a learning technique. Simple local color descriptors combined with the SIFT descriptors are used. Experiments using benchmark and complex generic object datasets are performed, and good performance is obtained. © 2008 Pleiades Publishing, Ltd.