一种改进的自适应关键帧提取方法

An Improved Algorithm of Adaptive Key Frame Extraction

针对关键帧提取方法中关键帧数目固定,不能根据镜头自适应调整的缺点,将帧间差法与运动变化信息熵方法相结合,提出一种自适应关键帧提取方法。根据镜头的变化确定提取的关键帧数目,采用运动信息的变化熵来提取关键帧,构建关键帧提取的数学模型,并通过实验进行验证。实验结果表明:该算法明显优于等间隔关键帧提取算法,提取的关键帧能较完整地表现序列图像的运动过程,更有利于对视频内容的理解。

Based on the problem that key frame number can not adaptively adjusted according to the lens, the paper proposed an adaptively key frame extraction method based on frame difference and the motion information entropy. The method determined the number of extracted key frame according to changes of the lens and extracted key frame based on the maximal motion and changing information entropy. A mathematical model for key frame extraction was constructed. And the theory of this method was introduced in detail. Many contrasting experiment have demonstrated that this method is obviously better than the key-fame extraction by equal interval. By this method the changing process of sequence is represented perfectly. It provides better perception and understanding for the video.