TheUseofPseudo-landmarksfor.PDFVIP

The Use of Pseudo-landmarks for Craniofacial Analysis: A Comparative Study with L1-Regularized Logistic Regression Ezgi Mercan1, Linda G. Shapiro1, Seth M. Weinberg2, Su-In Lee1 Abstract— Morphometrics, the quantitative analysis of shape, methods in the medical literature: 1) the well-known dense is used by craniofacial researchers to study abnormalities in correspondence method of Hammond and Hutton [5] and 2) human face shapes. Most of the work in craniofacial morpho- the more recent anthropometric mask approach of Claes [2], metrics uses landmark points that are manually marked on 3D face data and processed via a generalized Procrustes analysis. both of which require an initial set of hand-marked labels to For large data sets this manual process is very time-consuming. compute correspondences and have fairly heavy computation. Dense sets of pseudo-landmarks have also been proposed and We have developed a very simple, but effective, method successfully used for classi?cation and clustering, but the two that computes pseudo-landmarks by cutting through each 3D main methods in the literature are very computationally inten- head mesh with a set of horizontal planes and extracting a set sive. We have developed a computationally simple method that can compute pseudo-landmark points at different resolutions of points from each plane. Correspondences among heads are from 3D meshes of human faces. In this paper, we perform a not required, and the user does no hand marking. For classi- comparative study employing L1-regularized logistic regression ?cation purposes, we use L1-regularized logistic regression, to train a classi?er that predicts the sex of 500 normal adult which deals effectively with the high dimensionality of the face meshes in order to compare our method to two alternative data and modest number of samples. This method also allows pseudo-landmark methods and a distance matrix approach.Our results show