3D Representative Face and Clustering Based Illumination Estimation for Face Recognition and Expression Recognition

Abstract

Eliminating the negative effect caused by variant pose and illumination is a very critical problem for expression recognition. In this paper we propose a 3D representative face (RF) and clustering based method, which can estimate 13 illumination conditions under certain poses. First, all faces are adaptively categorized into 31 facial types by k-means clustering, so people with similar facial appearance are clustered together; Then the representative face of each cluster is generated. Finally we select the most discriminative features to train a group of SVM classifiers and get 96.88% estimation accuracy when estimating the test set with frontal view. Compared with other related works, ours does not rely on 3D reconstruction, and to get the generalization ability, we use our RF and clustering technique.