基于时空切片的关键帧提取研究

Key Frame Extraction Based on Space-time Slice

关键帧提取技术是视频检索和视频摘要等领域的关键技术之一。针对现有关键帧提取方法缺乏时空特性分析,而且提取结果不符合人眼视觉感知,为此提出了一种新的关键帧提取方法。首先提取视频镜头的时空切片,然后对视频切片进行K均值聚类,最后根据视频的时间连续性,在每个类中提取关键帧。大量实验表明,本方法能捕捉视频的主要内容,以极低的冗余性反映视频的动态特性,符合人眼视觉感知。

Key frame extraction is an important part in video retrieval and video abstraction.Existing methods are lack of spatio-temperal analysis and n ot consistent with human perception,therefore a new key frame extraction method was propos ed in this paper.Firstly,space-time slice of a shot is extracted,and then K-means clustering is made on the space-time slice,at last,key frames are extracted in every cluster according to the video temporal continuity.As the experimental results showed,this method can capture the main content and exhibit e xcellent visual continuity and low repetition,which is consistent with human perception.