Virtual human modeling from photographs for garment industry

摘要

The research presented in this paper is to develop a technique of virtual human modeling for the garment industry from two photographs of a human body in two orthogonal views. Firstly, an efficient segmentation method is applied on the two photographs to obtain the contours of the human body. After this, a template-based feature extraction algorithm is introduced to determine the feature points on the human contours by human morphology rules. Finally, a view-dependent deformation technique is described to construct the virtual human body by using human contours. Our segmentation algorithm is derived from the Mumford-Shah segmentation technology and the level set formulation, and it is accelerated by applying multi-pyramid levels. The deformation technique is related to axial deformation. With our deformation method, the reference silhouettes (the front-view and right-view silhouettes of the template human model) and the target silhouettes (the front-view and right-view silhouettes of the human body from the photographs) are used to deform the template human model, which is represented by a polygonal mesh with predefined features. The self-intersection problem in the axial deformation is solved in our deformation approach. Compared with other virtual human modeling approaches, the speed of constructing the human model is increased; and our deformation technique has better continuity and local deformation properties. At the end of the paper, some potential applications for the garment industry are given to demonstrate the functionality of virtual human models constructed by our approach.