A Synthesis of 3-D Kabuki Face from Ancient 2-D Images Using Multilevel Radial Basis Function

Abstract

A technique to synthesize a 3-D Kabuki face model from ancient images is proposed. Kabuki is a traditional Japanese theater art. Kabuki researchers can know some ancient information of Kabuki from the ancient 2-D images such as Ukiyo-e (Japanese drawing) and Kumadori images (Kabuki makeup). Compared with the real face images in photographs and paintings, the face images in Ukiyo-e and Kumadori images were distorted a little from the real face for art representation. Hence, it is necessary to synthesize a new 3-D face model between the 3-D real face original model and the 2-D images. A deformation technique is proposed for transforming 3-D real face original models according to the 2-D images, and a mapping algorithm is given to map the images onto the 3-D face shape model. To ensure balance between precision and smoothing of 3-D model deformations and mappings, the radial basis function (RBF) is developed and a multilevel RBF is proposed. This multilevel RBF solver can be used to deform the 3-D real face model as well as map the 2-D images onto the 3-D face model. Some experimental results are given to demonstrate the effect of this technique.