3D Face Recognition.pdf

Abstract—Recently, a 3D face recognition approach based on geometric invariant signatures, has been proposed. The key idea of the algorithm is a representation of the facial surface, invariant to isometric deformations, such as those resulting from facial expressions. One of the crucial stages in the construction of the geometric invariants is the measurement of geodesic distances on triangulated surfaces, carried out by fast marching on triangulated domains (FMTD). Proposed here is a method, which uses only the metric tensor of the surface for geodesic distance computation. When combined with photometric stereo used for facial surface acquisition, it allows constructing a bending- invariant representation of the face without reconstructing the 3D surface. Index Terms--face recognition, fast marching, photometric stereo, multidimensional scaling. I. INTRODUCTION ACE recognition is a biometric method that unlike other biometrics, is non-intrusive and can be used even without the subject’s knowledge. State-of-the-art face recognition systems are based on a 40-year heritage of 2D algorithms, dating back to the early 1960s [1]. The first face recognition methods used the geometry of key points (like the eyes, nose and mouth) and their geometric relationships (angles, lengths, ratios, etc.). In 1991, Turk and Pentland applied principal component analysis (PCA) to face imaging [2]. This has become known as the eigenface algorithm and is now a golden standard in face recognition. Later, algorithms inspired by eigenfaces that use similar ideas were proposed (see [3], [4], [5]). However, all the 2D (image-based) face recognition methods appear sensitive to illumination conditions, head orientations, facial expressions and makeup. These limitations of 2D methods stem directly from the limited information about the face contained in a 2D image. Recently, it became evident that the use of 3D data of the face can be of great help as 3D i