Low-Resolution Range Data Surface Matching for 3D Face Verification

Abstract

In this paper, we presented a novel approach for automatic 3D face verification by range data surface matching. The method consists of range data registration and comparison. There are three steps in registration procedure: the coarse step conducting the normalization by exploiting a priori knowledge of the human face and facial features to make faces have the similar attitude; the next step considering the existence of holes in the low-resolution range data will undermining the recognition results, we presented a novel approach for holes filling to improve the range data quality; and the fine step aligning the input data with the model in the database by the Delaunay-Iterative Closest Point (D-ICP) algorithm. During the face comparison, a Modified Hausdorff Distance (MHD) is employed as the similarity metrics. The experiments are carried out on a database with 30 individuals, and the best EER of 1.667% is achieved.