Cartoon Figure Generation Based On Matchingalgorithm

Cartoon figures of current animation are normally designed and drawn manually, which demand designers a high artistic design ability, and current computer animation design tools are only provide a interactive designing methods, all of the process from concepts to products are finished under the control of designers. What the final image will look like solely depends on the personal artistic sense and style.In this paper, we use simple stroke input and geometry shape as constraints, the features of stroke and shape are extracted and matched with the features of example graphics primitives in database, then generate the cartoon figure according to the user's intentions. The features of user's input script and graphics primitives in database are described by Fourier shape descriptor and chaining code descriptor, the similarity between shapes are measured by Euclidean distance, the modification of graphics primitives and generation of cartoon figure are done by interactive way. Our experiments show that the synthesis figure are consistent with the intention of users, and with the user's feedback, the result image will be more reasonable. In current works, the scale of segmentation of graphics primitives is not small enough to control and edit the appearance of cartoon figures, we will reduce the scale of graphics primitives to eye and mouth in the future work to express and control the feeling of figures more accurately.The first part of this dissertation is the background and meaning of our research work, the related works about cartoon generation are also introduce in this part. The idea of cartoon figure generation using shape matching are proposed in the second part, the process of the acquisition, classification and organization of example graphics primitive, the extraction and matching of features and the generation of cartoon figures are described in detail in this part. In the third section, the feature description method are introduced, we select Fourier descriptor as the feature descriptor, and the performance of it is shown the experiment results. Image feature matching methods are introduce in the fourth section, Euclidean distance is used as similarity measure and a sound cartoon figure are generated in our experiments. The research work are summarized in the fifth section, and the direction and perspective of our research are mentioned.