Shape analysis of female torsos based on discrete cosine transform

Abstract

Purpose – The purpose of this paper is to present results of shape analysis of female torso shape using the discrete cosine transform (DCT) from a three-dimensional (3D) whole body scan database. Design/methodology/approach – Torso shape is a central part of body shape and difficult to describe by linear measurements. In order to analyze body shape variation within a population the authors employed a DCT-based shape description method to compresses a dense 3D body scan surface into a small vector that preserves shape and removes size. The DCT-based shape descriptors of torso surfaces are further fed to principal component analysis (PCA) that decompose shape variation into constituent shape components. A visualization program was developed to observe principal components of torso shape and interpret their meanings. Findings – Extreme shapes of the first ten principal components summarize major shape variations and identify shapes that are difficult to capture with traditional anthropometric measurements. PCA results also help to find and retrieve similar shapes from a population-level database. Originality/value – Using the DCT for PCA of torso shape is a unique and original approach. It provides a basis for the description and classification of torso shape in 3D and the results from the shape analysis are potentially useful for designers of clothing and personal protective equipment.