Research On Key Techniques Of Video-Animation’s Material Retrieval

Animation creation is very complicated and needs many cartoonists to draw constantly. We also know much available information coming from the real world. If we can transform the information in the real world with the technology of image processing to animation, it is very meaningful for animation creation.Recent years, many experts propose lots of algorithms to realize this idea. There has been a lot of achievement for the idea. For example, the “Facial cartoon automatic generation system” and “Color sketch technique” developed by Microsoft Research Asia have been used “Digital greeting card making system” in Japan.As the research content of video-animation, in this paper, we study techniques of video-animation to save, retrieve for creation efficiency of animation. The main work in the paper is :(1) In order to search for the shot related to particular role quickly and use the materials that have existed efficiently, we design and develop shot segmentation algorithm that can extract the shoots and the key-frames in the videos automatically. Sametime, the system provides interfaces to invoking the algorithm to transform the image to cartoon image for users. Then we can extract the feature including face, text, content feature from key-frames and cartoon image with image processing and face recognition technology. Finally, we provide more ways of image retrieval with text retrieval, image retrieval and face retrieval to help creators retrieve the related shots and materials.(2) The study of hypergraph-based image retrieval algorithm. Hypergraph describe similarity relations of multiple images. Firstly we narrow the scope of image retrieval by clustering. Then we improve the speed of retrieval based the result of clustering in the first step.(3) By comparing the K-Means with Spectral Clusteing with Hypergraph Clustering, Hypergraph with Mainfold Ranking and Linear Ranking and the result before clutering with the result after clustering through experiments, the result of experiments shows: hypergraph clustering can improves the precision of clustering; hypergraph ranking based the result of hypergraph clustering can improve the precision ration of image retrieval and reduce time cost.