Isoradius Contours: New Representations and Techniques for 3D Face Registration and Matching

Abstract

We propose a technique for 3D face registration and matching using a novel representation called "isoradius contours". An isoradius contour is the contour on the 3D facial surface that is a known fixed distance relative to some predefined reference point (the tip of the nose). A 3D face representation contains many isoradius contours with different radii and the first major benefit of the technique is that the shape of the contours is independent of the facial pose, due to the infinite rotational symmetry of a sphere. The second major benefit of the technique is that registration, alignment and matching can be implemented using a simple process of ID correlation. Our results have shown that registration and alignment is of comparable accuracy to ICP (iterative closest points), but is fast, non iterative, and is robust to the presence of outliers.