Automatic 3D Garment Fitting Based on Skeleton Driving

Abstract

In this paper, we propose an automated 3D garment fitting method. The proposed method can automatically position a 3D garment model onto a human model without any user manual intervention. Given a garment model and a human model with various postures and shapes, we firstly segment the garment and the human model separately into mutually corresponding parts by our improved segmentation method. Secondly, we extract the skeletons of these parts accurately according to the segmentation. According to the skeletal information, we analyze the posture difference between the human and garment models. Thirdly, we change the posture of the garment model according to the difference of skeletal posture. Fourthly, we position the garment model onto the human model roughly with feature points of skeletons. Lastly, we constrain the garment mesh to resolve the penetration between human and garment models. With the purpose of achieving more realistic dressing result, we use mass-spring system to simulate garment. The experimental results show that the proposed method is stable and efficient.