A Method of Regional High Level Feature Extraction on Video
一种视频局部高层语义特征提取算法

Extracting high level features from video accurately benefits the content-based video retrieval. Regional high level feature describes the object in frames. Considering the characteristic of objects and its specific scene, we propose a method of regional high level feature extraction on video combining the global information and regional information of frames together. The experiment based on the TRECVID2005 data set illustrates that this method has a better performance than other methods which use regional information and global information separately.