Three Dimensional Cartoon Animation Based On Art Rules

Three-dimensional cartoon facial animation, which is a new branch of the field of non-photorealistic rendering, has become a hot area of the computer graphics. With the fast development of the animative film and television, video games, virtual meetings, research in this field increased gradually. In this thesis, a brief revision of the research in the field of the cartoon facial animation is first presented along with both advantages and disadvantages of each method, after which, methods developed in this thesis in the three key parts of the cartoon facial animation are described in detail. The contributions of this thesis are summarized as follows:Firstly, a kind of three dimensional reconstruction method based on aesthetic rules is proposed. In this method, general exaggeration techniques of the artists are introduced for the exaggeration of the model. The exaggerations are performed after the detection of the parts in the human face that are usually exaggerated by the artists. For improving the accuracy of the feature point extracted, the feature points in the forehead are determined with other points detected with the ASM algorithm by interpolation.Secondly, a texture cartoon method is proposed for a single photo. In this method, the face of a particular person is aligned to the cartoon photo template and then fused to obtained cartoon version of the particular face. The cartoon extent of the particular face can be adjusted by the adjustment of the fusion coefficient. The face region is firstly fused to the background with the Poisson equation. For the acceleration of the face and background fusion process, a fast texture interpolation method is proposed in which the pixel in the skin corresponding to each pixel in the background is pre-computed according to the cartoon template. Experimental results show that satisfied texture fusion results can be achieved with high speed using the proposed texture interpolation method.Finally, a 3D animation method is proposed based on the non-specific model parametric conversion where the animation parameters suitable for the model in this thesis are converted from a reference model. The animation parameters of a vertex in the model of this thesis are computed from the position relationship between the vertex in our model and the three vertices of the triangular element in the reference model. This method can effectively avoid the zero problem of the animation parameters in the traditional methods. Experimental results show that this method is more plausible and can satisfy the actual application requirement with natural cartoon animation effect.