改进的蚁群算法与凝聚相结合的关键帧提取

Key frame extraction based on improved ant algorithm and agglomerative

关键帧提取技术,对基于内容的视频检索有着重要的作用。为了从不同类型的视频中有效地提取关键帧,提出了改进的蚁群算法与凝聚相结合的关键帧提取算法。该方法提取视频中每帧的颜色与边缘特征向量,利用改进的蚁群算法自组织地对颜色和边缘特征向量进行聚类,得到初始聚类。通过凝聚算法对初始聚类进行优化,得到最终聚类。提取每类中距离聚类中心最近的向量,将其对应帧作为关键帧。实验结果表明:使用该算法提取的关键帧不仅可以充分表达出视频的主要内容,而且可以根据视频内容的变化提取出适当数量的关键帧。

The key frame extraction is very important to content-based video retrieval. In order to extract key frame efficiently from different types of video, an efficient method for key frame extraction based on improved ant algorithm and agglomerative is proposed. An improved ant algorithm is applied to the histogram differences and texture of video shot self-organized, and obtains an initial clustering result; agglomerative is conducted to optimize the initial clustering result and then a final clustering result is obtained; the center frame of each clustering is extracted as the key frame. The experiment result shows that, the key frames extracted by using this algorithm can adequately express the primary content of the video, and proper quantities of key frames also can be extracted according to the change of the video content.