Abstract: A new technique is presented for modeling personal virtual clothed human model with images in orthogonal views.The modeling process is template-based.Firstly,a method for constructing a cross-section based template human model from a scanning human model is proposed.A method for describing the intrinsic relationship between cross-sections and the 3D skeleton of template human model is proposed.Secondly,2D skeletons of a human in images are set up after automatically extracting feature points on human contours.Thirdly,the relationship between the cross-sections and images is established with the correspondence between 3D skeleton of the template human model and 2D skeletons on images.Finally,a personal virtual human model is obtained by parametrically deforming cross-sections with shape parameters that is automatically gotten from images.Comparing with other methods,our approach is feature and parameterization based,which makes human model deformation more convenient and efficient;the mapping uncertainty between vertices on 3D human model and 2D images is removed;and self-intersection during human model deformation is also avoided.