Accurate skin deformation model of forearm using MRI

This paper presents a new methodology for constructing a skin deformation model using MRI and generating accurate skin deformations based on the model. Many methods to generate skin deformations have been proposed and they are classified into three main types. The first type is anatomically based modeling. Anatomically accurate deformations can be reconstructed but computation time is long and controlling generated motion is difficult. In addition, modeling whole body is very difficult. The second is skeleton-subspace deformation (SSD). SSD is easy to implement and fast to compute so it is the most common technique today. However, accurate skin deformations can't be easily realized with SSD. The last type consists of data-driven approaches including example-based methods. In order to construct our model from MRI images, we employ an example-based method. Using examples obtained from medical images, skin deformations can be modeled related to skeleton motions. Retargeting generated motions to other characters is generally difficult with this kind of methods. Kurihara and Miyata realize accurate skin deformations from CT images [Kurihara et al. 2004], but it doesn't mention the possibility of retargeting. With our model, however, generated deformations can be retargeted. Once the model is constructed, accurate skin deformations are easily generated applying our model to a skin mesh. In our experiment, we construct a skin deformation model which reconstructs pronosupination, rotational movement of forearm, and we use range scan data as a skin mesh to apply our model and generate accurate skin deformations.