Improved active shape model for facial feature localization using poem descriptor

Abstract

Aiming at the problem in facial feature location, the Active Shape Models algorithm suffers from variations of poses, illumination and expressions. In this paper, we first introduce the patterns of oriented edge magnitudes descriptor to represent the local appearances of landmarks. It can provide more robust and accurate guidance for search than grey-level profiles. Moreover, the vector of shape parameters is improved by defining the rotation factor. Experimental results show that the proposed method significantly outperforms the original ASM and ASM+LBP method under attitude and illumination variation.