With the development of multimedia and computer network technique, the volume of digital image and video is going up fabulously. It has a good sense to organize, manage and utilize the data of the image and video efficiently which makes the content-based image and video retrieval be on the top research of the multimedia area. But there still are a lot of problems needed to be resolved in the content-based image and video retrieval at present.In this paper, we develop two prototype systems started with the accuracy and time of the content-based image and video retrieval. One of them is a content-based image retrieval, the other is a semantic-based news video retrieval system. Relevance Feedback and Support Vector Machine learning are introduced into the contend-based image retrieval system. We also compare DTW&GSVM results to UFM&GSVM's which proves that the algorithm of DTW&GSVM is effective. The first shot segmentation and the second shot segmentation used in the semantic-based news video retrieval system make the shots segmentation accurate;then, based on this, we introduce interactive of news video scene segmentation. The semantic retrieval of news video is implemented by the annotation of scene through metadata annotation.There are 5 sections in this paper. In section l,both the job we have done and the research state of content-based image and video retrieval are described. Section 2 introduces the knowledge of pattern recognition and Support Vector Machine. The design and realization of a region-based image retrieval prototype system is provided in section 3. Section 4 describes the design and realization of a semantic-based news video retrieval system. And finally, we conclude in section 5,together with a discussion of future work. |