Statistics for 3D Human Body Forms

Abstract

A method of calculating representative forms from a given set of forms was developed, in which surface data is modeled by polygons based on landmarks. Inter-individual distances are defined as distortions in FFD control points. By calculating inter-individual distances for all possible pairs of forms, the distribution of the 3D forms in m-dimensional space is obtained using MDS. Each MDS dimension represents an independent shape factor. Forms with specific MDS scores, such as (0.5,0,0,0), (1,0,0,0) in standard deviation units, are calculated as weighted averages of all actual forms. An FFD transformation grid is calculated that represents the systematic form transformation along an MDS dimension. Forms with different scores for the first or second MDS dimensions only and average scores (=0) for the other MDS dimensions are calculated using these transformation grids.