Predicting Realistic and Precise Human Body Models Under Clothing Based on Orthogonal-view Photos

Abstract

Accurate and realistic digital human body models are required by many research applications, for example in the areas of ergonomics, clothing technology, and computer graphics. The already difficult research problem becomes more challenging if the individual subjects to be modelled are dressed in normal or loose-fit clothing. In this study, we present an intelligent two-phase method to customize 3D digital human body models based on two orthogonal-view photos of the customers. It integrates both image-based and example-based modelling techniques to create human body models for individual customers with precise body measurements and realistic appearance. It fills up the research gap of human model customization; without the need of taking body scan, any customers can create their 3D digital body models only based on their orthogonal-view photos in normal or loose-fit clothing. Experimental results have shown that the proposed method can efficiently and accurately customize human models of diverse shapes, meeting the specific needs of the clothing industry.