3D Facial Normalization with Spin Images and Influence of Range Data Calculation over Face Verification

In this paper face verification techniques have been performed over 3D data acquired by a Laser Scanner. Advantages of 3D face models have been used to perform a normalization task over each face. First, an original method to detect local features points, based on Spin Images, has been developed. Once local features, as the nose tip or eyes corners, have been detected, a normalization process is carried out. After face normalization, different depth maps are calculated using several transform functions to equalize the images. The adequacy of each equalization to face verification has been measured to determine which one emphasizes most the feature discrimination. Face verification has been performed through a Principal Component Analysis and a Support Vector Machine. Final results show the importance of a careful 3D normalization and an optimal election of the depth map towards a improvement in the verification method.