3D Head Modeling Based On Different Gender Head Mesh Model

With the development of virtual reality technology, which is originally used in many fields of people’s daily life, the research of personalized virtual human is becoming hotter and hotter. Head is the most important part in performing personality of the virtual human. Research on3D head modeling of virtual human has lasted for decades. Many researchers use3D scanning devices or modeling software to obtain vivid3D head model, but these methods have some limitations in terms of economy and operability. Constructing a model with geometric and topological structures is the focus of this paper, certainly it also should be realistic and easy to control.And we should make the model realistic and valuable.Aiming at the problems that the model can’t express the topology of the head completely or the model is too complicated, firstly, we build a new general head grid model by combining the Candide-3model with the simplified mesh model. And after adjusting the model with a certain proportion, we can get the mesh models with different sexes. Secondly, the method of ASM (Active Shape Model) is proposed to automatically extract the feature points when we input the full-face photo. The Euclidean distances between certain feature points are calculated from the output of coordinate information of extracted points. The sex is recognized through the comparisons with statistical results of the input images. Thirdly, we obtain the matched model through the deformations of the feature and non-feature points. Then for the synthetic model, we use an adaptive texture mapping method to get the reconstructed model and adopt the Image Quilting method to solve the problem of texture seems in a certain extent. Lastly, we synthesize six basic facial expressions with AU. The experimental results show that the new synthesized model can construct the topology of mesh model well and can be controlled easily. We can promote the speed of reconstruction in a certain extent through selecting corresponding gender model. And by using adaptive texture mapping, we get a better sense of the head reconstruction. In summary the methods proposed in this paper can prove some adaptability and application value.