Dressing virtual humans from 3D scanned data

This paper presents a new approach to dressing three-dimensional virtual human models from scanned data. As the first step in this process, an algorithm is developed to obtain key body features and a reference skeleton. Due to the difference in body shape among various people, especially different ethnicities, we propose a two-step method for finding body feature points. Firstly, body feature regions are located through some pre-defined parameters defined by the proportion of head to body height. We adopt the idea of describing the body height as eight-heads tall. Secondly, body features are extracted within their corresponding regions. In the second step, garment patterns are constructed based on body features and the reference skeleton described above using two methods. One is a redial offset method the other is a swept surface method.