Feature detection and tracking on geometric mesh data using a combined global and local shape model for face analysis

Automatic geometric feature localization is the first step towards the 3D based face analysis. In this paper we propose a shape model with a local and global constraint for feature detection. Such a so-called shape-index based statistical shape model (SI-SSM) makes use of the global shape of the facial data as well as local patches, consisting of shape index values, around landmark features. The fitting process and the performance of our proposed method are evaluated in terms of various imaging conditions and data qualities. The efficacy of the detected landmarks is validated through applications for geometric based face identification.