一种基于核聚类的关键帧提取方法

A Key Frame Extraction Algorithm Based on Kernel Clustering

为了在视频数据库中提供有效的视频检索和浏览功能,必须用简明的方式表示视频的内容。关键帧是对视频镜头的简洁表示,关键帧提取已成为视频检索的一个重要研究方向。文中提出了一种基于核聚类的视频关键帧提取方法,它通过对视频提取颜色特征,并将这些特征作为样本映射到高维特征空间之后,在特征空间中进行聚类,使原来没有显现的特征突现出来,自动将内容相似的样本归为同类,每一类可取一个样本代表其内容,这样的样本即为关键帧。实验结果表明这种方法可以较好地概括视频内容。

To provide effective video browse, search and retrieval ability, video content should be summarized with compact but proper representation. Key frame is a simple effective form of summarizing a long video sequence. And key frame extraction has been recognized as one of the important research issues in video information retrieval. A key frame extraction algorithm based on kernel clustering is proposed in this paper. This method extracts the character of video, and maps the character in the original space to a high-dimensional feature space which we can perform clustering efficiently. We can extract a key frame from each cluster. Computer simulations show this algorithm can give a good representation of video content and make video retrieval much easier.