Automatic design-preserving virtual garment transfer

Abstract

Design-preserving garment transfer which can transfer a garment from one body to another is a powerful technique in computer graphics. In this paper, we propose a novel method for automatic design-preserving garment transfer. Firstly, the correspondence between the source garment and the source body is automatically calculated by finding the nearest face’s barycenter from the garment to the body. Secondly, the mesh deformation transfer algorithm is utilized to obtain the target garment. After aligning the target garment and body with the ICP algorithm, the virtual fitting result can finally be obtained by fitting the target garment onto the target body. For a new garment worn on the source body, our approach can quickly and effectively transfer the garment to the target body while preserving the designed details. Two experiments were carried out to validate the applicability and effectiveness of our method. The results showed that our method could be applied to different garments for human bodies with different shapes and poses. Additionally, we introduced a new method for evaluating our transferred garments based on similarity and fit. Moreover, our approach can be applied to garments with different particle distances and performs better than the ‘Auto grading’ technique of CLO 3D. Our proposed method is simple, effective, automated and fast. Therefore, it has a significant potential for improving personalized garment development efficiency.