A system for producing an accurate human body model is important and required in many fields. This study describes a complete system for creating an anatomically accurate 3D human body model from living human scans of a specific person. The method provides accurate and personalized tissue deformation arising from articulated motions of human body regions such as the human hand, probably the most complex articulated region in the human body.;Our system approaches the problem in three stages. First, a person-specific hand model is generated from a single canonically posed palm image of the hand without human guidance. Tensor voting is employed to extract the principal creases on the palmar surface. Joint locations are estimated using extracted features and analysis of surface anatomy. The skin geometry of a generic 3D hand model is deformed using radial basis functions guided by correspondences to the extracted surface anatomy and hand contours. The result is a 3D model of an individual's hand, with similar joint locations, contours, and skin texture.;Second, a parallel skin deformation algorithm is presented suitable for SIMD (Single Instruction Multiple Data) architectures such as GPUs. It presents real-time performance of skin deformation arising from the motion. Joint weights of each vertex are automatically computed and used in skinning algorithms. Skin deformation algorithms for articulated subject such as SSD (Skeletal Subspace Deformation), PSD (Pose Space Deformation), and WPSD (Weighted Pose Space Deformation) are parallelized in SIMD manner and implemented on GPU fragment processors to obtain sufficient speed-up for real-time skin deformation.;Finally, a complete system is developed to create a deformable articulated human body volume from multiple 3D MRI scans of an in vivo human body, which can produce anatomically accurate human volume deformation for articulated body animation. The system combines existing and new techniques that together address the practical issues involved in producing a detailed articulated volume model from scans. These include joint estimation from the actual skeleton, bone volume registration, smooth deformation that preserves the rigidity of the skeleton, and a staged registration process that incorporates a careful initialization as needed to deal with the strong local minima inherent in registering articulated bodies. A locally adaptive non-rigid registration algorithm is developed that greatly reduces the high degrees of freedom and data processing arising in registration of deformable articulated volumes. The registration step establishes a correspondence across scans that permits the last step, interpolation of the scans under intuitive joint control using an example based deformation approach that we extend from the surface to the volume domain.;The result is a volumetric model driven by standard joint control that incorporates detailed deformation based on real data. Visual presentation and quantification of articulated human body model and deformation have potential to contribute many applications such as medical studies of arthritis, bio-mechanics, robotics, ergonomics, medical education, rehabilitation, virtual reality as well as computer graphics and animation.
展开▼
