摘要
为了有效、自适应地提取镜头中的关键帧,提出了一种基于吞噬聚类的关键帧提取新算法。该算法通过邻近吞噬体的互相吞噬,聚类相似的数据对象,最后所剩吞噬体的吞噬中心即为视频帧的聚类中心,距离聚类中心最近的特征向量所代表的视频帧就是所需要的镜头关键帧。对该算法利用MATLAB仿真并通过与传统算法对比,结果表明,相对于传统算法,该算法的查全率和查准率都有了一定程度的提高。
In order to extract the key frames in the shot self-adaptively and effectively, a new algorithm of key frame extraction based on swallowing cluster is proposed. This algorithm gathers similar data objects through swallowing each other between the adjacent swallowing objects, and the swallowing centers of the swallowing objects leaved behind at last are the cluster centers of the frames. Frames represented by the features vectors which are the nearest from the cluster centers are the key frames. Simulating the new arithmetic through the MATLAB, and comparing it with the traditional arithmetic, it shows that the recall ratio and precision ratio have been improved to some extent.
出处
《电视技术》
北大核心
2014年第13期212-214,共3页
Video Engineering
关键词
视频检索
关键帧提取
吞噬聚类算法
video retrieval
key frame extraction
algorithm of swallowing cluster