摘要
通过分块的色彩直方图对视频进行分割,提取每个视频子片段的关键帧,并提出了一种以SIFT特征为基础的语义词袋方法实现视频指纹的提取,该方法既提高了视频指纹提取的速度,又能得到具有较高独立性和鲁棒性的视频指纹。利用串匹配的方法实现视频指纹的匹配检测,并进一步对相似度进行加权处理,提高了匹配效率与匹配精度。
A novel scheme is proposed in this paper to retrieve videos based on video fingerprint. Videos are separated into several segments through block-based color histogram. Then video/ingerprint is extracted by using semantic word bags based on SIFT features, which can not only improve the video fingerprint extraction speed, but also can achieve video fingerprint with high independence and robustness. A string matching method is employed to implement video fingerprint matching. What＇ s more, the processing of similarity weighting can also make a great improvement of matching efficiency and matching accuracy.
出处
《电视技术》
北大核心
2014年第15期16-19,28,共5页
Video Engineering
基金
广电总局科技司项目
关键词
视频检索
关键帧
视频指纹
特征匹配
video retrieval
key frame
video fingerprint
features matching
