Method of facial feature positioning based on improved ASM

Abstract

Facial Feature Positioning is attracting many attentions in computer vision. Active Shape Model (ASM) is considered as high level image processing algorithm, especially in face alignment. Active shape model is used to model the change of face shape using principal component analysis (PCA). Because of the influence of posture and expression, the linear PCA model is difficult to describe the face shape changes under different expressions and gestures. This paper presents a method based on conventional ASM. The proposed method exploits the Scale Invariant Feature Transform (SIFT) descriptors and uses AdaBoost algorithm to detect the face area to make sure the initial shape model at a perfect position. Experiment results show that the proposed method performs well in localizing facial feature than original ASM.