Sphere-spin-image: A viewpoint-invariant surface representation for 3D face recognition

Abstract

This paper presents a new free-form surface representation scheme, which we call Sphere-Spin-Image(SSI), and its application to 3D face recognition. An SSI, associated with a point on the surface, is a 2D histogram constructed from a neighborhood surface of the point using position information, which captures the characteristic of local shape. Thus, a free-form surface can be represented by a series of SSIs. Correlation coefficient is used as similarity metric for comparing SSIs. During face recognition, the SSIs of points on face surface are computed, and recognition task is achieved by SSI-comparison-based voting method. To reduce computational cost, only some particular points of face surface are involved in voting. With face database consisting of 31 different pose models, experimental result of equal error rate 8.32% demonstrates performance of the proposed method. © Springer-Verlag Berlin Heidelberg 2004.