A Novel Video Annotation Framework Based on Video Object

Video annotation is very important for video management, such as video retrieval. Despite continuous efforts in inventing new annotation algorithms, the annotation performance is usually unsatisfactory, and the annotation vocabulary is still limited due to the use of a small scale training set. In this paper, a novel video annotation framework based on the video object is presented, named Object-Based Video Annotation. By dividing video into three types, we deal with different kind of video in different way. The first kind of video was annotated by human base on the e-Annotation architecture. The second kind of video was automatically annotated by the web mining methods. The third kind of vi eo annotated by video analysis model which detect the video object and label them at the same time. Then active learning model implement active learning method in the video database, which can add new labels and video in the database. We also present an application system based on annotations: video retrieval. At the same time we add relevance feedback in our framework to optimize the result. The system designed base on a real-world situation by including video gathered from the Internet and is designed for exploratory video retrieval system based on the internet.