Classification and sampling of shapes through semiparametric Skew-Symmetric Shape Model

We present a novel method for shape modeling using an extended class of semiparametric skew-symmetric (SSS) distributions. Given several realizations of a simple shape, the proposed method models it as a joint distribution of angle and distance from the centroid of all points on the boundary. The model, called “Semiparametric Skew-Symmetric Shape Model” (SSSM), is capable of capturing inherent variability of shapes provided the realization contours remain within a certain neighborhood range around a “mean” with high probability. In this paper, we will discuss SSSM learning, classification through SSSM and sampling shapes from the learned models.