一种两阶段的关键帧提取方法

摘要

The keyframe extraction algorithm based on co-hisgram leads to the problem of missing extraction, while the method based on edge feature leads to the problem of high computational complex. The article proposes a method that takes advantages of the two algorithms, namely, first extracts key frames by using co-hisgram, then extracts the finally key frames from the previous results. Experimental results demonstrate that frames extracted can give a good summarization of the video while avoid redundancy.%基于联合直方图的帧间相似性在关键帧提取中存在漏检问题,而基于边缘特征的方法,虽然效果较好,但在计算复杂度方面比较高.本文结合这两种方法的优点进行两阶段的提取方法,即先用基于联合直方图的方法提取视频的候选关键帧,再用基于边缘特征的方法进行二次提取.实验结果表明,该方法提取出来的关键帧有较好的代表性,冗余度底.