Animating scanned human models

Abstract

We present techniques for automatically creating and animating models obtained from human whole body scanned data. A layered model is developed in which the underlying skeleton, simplified surface and mapping of the surface to the skeletal structure are generated without manual intervention. External body landmarks such as those employed in anthropometric surveys are used to obtain the skeleton, these landmarks are then reused to help map the model surface consistently to the skeleton. The main contribution of this work is to show how automatically or even manually extracted surface landmarks greatly simplify the task of skeleton generation and surface bone mapping, the later uses surface connectivity to grow surface regions as opposed to conventional volume searches. The techniques are illustrated by animation from motion capture data in which a vertex blending technique is used to create continuous and smooth deformation.