Texture Mapping Of Realistic 3D Facial Models

The reconstruction and texture mapping of realistic human facial models is always attractive to researchers in computer graphics, and has been studied for many years. With the development of computer graphics technique, applications of reconstruction and texture mapping include such diverse fields as character animation for films and advertising, computer games, virtual reality, teleconferencing, user-interface, surgical facial planning, and criminal discrimination etc.People always set up models to imitate real objective world, but the real world is one of the most complex, objects to model in computer graphics, especially the imperceptible configuration of the true world. Modeling of all these models is not only hard to establish, the calculation quantity is huge, but also request a large storing space. Hard to satisfy the request to show in real-time, people put forward texture mapping technique. Texture mapping is one of the most successful new techniques in high quality images synthesis. Its use can enhance the visual richness of rester scan images immensely while entailing only a relatively small increase in computation.Texture mapping enhances the visual realism of 3D models by adding fine details. To achieve the best results, it is often necessary to force a correspondence between some of the details of the texture and the features of the model. Interactive 3D authoring applications frequently map 2D textures onto a 3D triangle mesh. This involves assigning (u,v) coordinates to the vertices of the 3D mesh, which defines a piecewise affine mapping of the texture image to the 3D mesh triangles. For example, when texture mapping a photograph of a face to a 3D mesh of a head, it is important that the eyes and the nose in the photograph are aligned precisely with the eyes and the nose in the mesh. In 1974, Catmull introduced texture mapping, he defined (u,v) as a real bivariate domain(texture domain) and (,s,t) expressed the parameter of curved surface domain. The problem of texture mapping can be stated as the following: for a given curved surface and a texture domain, define a mapping between these two domains. After the study of Catmull, many researchers have done a lot of work towards the improvement of the quality of the generated images and texture mapping.Texture mapping is widely used to the realistic images. In the paper we introduce the definitions on texture, texture mapping and the least squares conformal maps for texture mapping etc.The 3D facial model used in this paper is based on two photographs of people's faces. This method can be applicable to general 2D models to 3D models. By establishing a one to one correspondence of the texture obtained with the use of camera to the points on 3D mesh models, carry on the realistic texture mapping and generate a realistic 3D facial model. By using this method, one may realize the reappearance of different faces of people, and the method can be used to create a 3D digital camera. The method can also be used to restore simple object models. This technique will be widely applied in many fields.In this paper, the 3D models are obtained by acquiring point cloud of 3D person's face according to two photos of a person's face. The main method used is to search initial points from two photos taking in different sides, generate some corresponding points with the use of a generating algorithm and finally obtain the coordinates of the 3D points cloud of the person's face by combining the camera parameters and the corresponding points of the photos.The models used in texture mapping is the triangle mesh generated by connecting 3D point cloud. The main idea of the algorithm is to lit a plane that contains all points in 3D point cloud with the method of least square, that is, the least square plane. Rotate the least square plane to a. plane which is parallel to the xy plane, then the 3D model is also rotated. The parameterizing of 3D point cloud will be finished by projecting the 3D model to the least square plane. The set of 2D point cloud is hence obtained. The 3D point cloud triangle mesh is then obtained by triangulating 2D points and mapping the parameterized 2D point cloud and the connecting relationship to 3D point cloud.To get the mapping from 2D texture to 3D surface, firstly one should establish a corresponding relating between the 3D surface and the points on the 2D plane, i.e., to find the corresponding relating between the texture plane and the 3D surface. After all steps as above, one may form a one-to-one mapping from the texture and a generated image. With the use of this map, one may map the texture to a 3D model The final step is to show a realistic image on a screen which is the result of the texture mapping. To emphasize the visional reality, we map a person's photo as a texture to a 3D model and obtain a good result, which realize the reconstruction of 3D facial model.Moreover, this paper also carries out a conformal texture mapping. A mapping X is called a conformal mapping if for all point (u,v), we havewhere N(u. v) denotes the unit normal to the surface.With this method, we apply least square method to the triangle mesh model formed from the date obtained with a 3D scanner, parameterize the mesh to a plane, adjust the texture and map the texture to the plane, then restore the 3D model by using the connecting relation of the parameterized plane and achieve a realistic model. We may see from this result that this texture mapping is almost conformal.Since the study of the technique in treating 3D graphes and images is still inadequate, many further work should be done. At the end of this paper, we sum up some aspects which may be improved and suggest some new subjects which may be studied further.The work in this paper provide some effective method in studying the texture mapping of facial models. However, some further work has to be done. The result of texture mapping obtained with the use of the method in this paper add some character and details to the 3D facial models. One should avoid aliasing and distorting at the places such as mouthes and noses where the curvature changes greatly. On the other hand, currently, we have only studied the gypsum models, the study on real person should be further developed. Finally, the texture of the gypsum models used in this paper was taken at special point sides, the texture of real person used at the end of this paper has been treated with some extra work. The algorithm we are using may not give perfect mapping results for every texture. The further study should pay special attention to character extraction and matching algorithm with intelligence. We hope that one may generate realistic 3D facial models automatically and in real time by raise the operation speed and. improve the accuracy of the algorithm. At the same time, one may give maps between any 3D models and any textures simply and instantly and hence build automatic 3D digital cameras.