基于光流方向信息熵统计的微表情捕捉

发布时间：2018-11-25 08:59

【摘要】：以光流法为依据,提出了一种基于光流方向信息熵(entropy of oriented optical flow,EOF)统计的方法捕捉微表情关键帧.首先,采用改进的Horn-Schunck光流法提取视频流中相邻两帧图像的微表情运动特征;其次,采用阈值分析法筛选出投影速度模值较大的光流向量;之后,采用图像信息熵统计光流变化角度,进而得到视频序列的方向信息熵向量,通过对熵向量的分析,实现微表情关键帧捕捉;最后,本实验采用芬兰奥卢大学的SMIC微表情数据库和中国科学院心理研究所傅小兰的CASME微表情数据库作为实验样本,通过与传统的帧差法比较,证明了本文提出的算法优于帧差法,能够较好地表现出微表情变化趋势,为微表情识别提供基础.
[Abstract]:Based on the optical flow method, a new method based on the direction entropy of optical flow (entropy of oriented optical flow,EOF) is proposed to capture the key frame of micro-expression. Firstly, the improved Horn-Schunck optical flow method is used to extract the micro-expression motion features of the adjacent two frames in the video stream. Secondly, the threshold analysis method is used to screen out the optical flow vectors with large projective velocity modules. Then, the direction information entropy vector of video sequence is obtained by using image information entropy to calculate the angle of optical flow change, and the key frame capture of micro-expression is realized by analyzing the entropy vector. Finally, the SMIC microfacial expression database of the University of Oulu, Finland, and the CASME microfacial expression database of Fu Xiaolan, Institute of Psychology, Chinese Academy of Sciences, were used as the experimental samples, and compared with the traditional frame difference method. It is proved that the algorithm proposed in this paper is superior to the frame difference method and can well show the trend of microfacial expression and provide the basis for micro-expression recognition.
【作者单位】：北京科技大学计算机与通信工程学院;合肥工业大学计算机与信息学院;
【基金】：国家自然科学基金资助项目(61672093,61432004) 国家重点研发计划重点专项课题资助项目(2016YFB1001404)
【分类号】：TP391.41

【相似文献】