互信息熵和Prewitt差测度的Lasso模型关键帧提取  

Lasso model key frame extraction for mutual information entropy and Prewitt difference measure

作　　者：高永 郝晓丽[1] 吕进来[1] GAO Yong;HAO Xiaoli;LU Jinlai(College of Computer Science and Technology , Taiyuan University of Technology , Jinzhong, Shanxi 030600, Chin)

机构地区：[1]太原理工大学计算机科学与技术学院,山西晋中030600

出　　处：《中国科技论文》2017年第20期2342-2348,2354,共8页China Sciencepaper

基　　金：山西省科技攻关项目(20130321007-02)

摘　　要：针对视频关键帧提取过程中,需要对视频中每帧图像进行大量诸如特征提取、图像配准、冗余消除等高复杂度的计算,占用大量的运算时间等问题。提出了1种互信息熵和Prewitt差测度的Lasso模型关键帧提取算法。该算法首先利用视频分割技术,将视频按照镜头变化分割得到了视频片段;其次利用互信息熵和Prewitt双重特征量提取视频片段中的关键帧并采用边缘匹配算法消除冗余帧序列。最后,得到的关键帧序列用Lasso模型检验收敛性。理论分析和实验结果表明,与基于加权的多视图关键帧提取算法相比,关键帧提取时间平均缩短了28.7s,验证了本文算法的高效性。In the process of video key frame extraction, it is necessary to perform a lot of such as feature extraction, image registration and redundancy elimination, In this paper, we propose a key framrithm for Lasso model based on mutual information entropy and Prewitt difference measure.used to segment the video objects. Secondly, the key frame in the video segment is extracted by mutual entropy and the Prewittdouble feature, and the edge matching algorithm s used to eliminate the redundant frame sequence. Finally, the resulting keyframe sequence is tested by the Lasso modll for convergence. Theoretical analysis and experimental with the weighted multi-view key frame extraction algorithm, the average extraction time of key frames is shortened by 28. 7 s,which verifies the efficiency of the proposed algorithm.

关 键 词：互信息熵 关键帧提取 PREWITT算子 边缘匹配 Lasso回归 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]