摘要
在视频流环境下,提出了一种融合序列蒙特卡罗滤波(SMCF)与加权直方图的快速人脸匹配方法,该方法以目标人脸在HSV空间下的加权直方图作为匹配特征,利用SMCF的采样和权值化处理技术预测人脸可能的匹配区域,通过最大后验估计实现目标的快速相似匹配。实验表明,采用在HSV下具有聚类特性的加权直方图特征可有效地适应复杂背景的干扰,同时SMCF的区域预测和最大化估计有效地减小了计算开销,增加了匹配的可靠性,可应用于人脸跟踪、视频监控及数字娱乐等领域。
This paper presents a new method to match human face with sequential Monte Carlo filter and weighted histogram in the context of streaming video.In this method the weighted histogram of target face in HSV is adopted as matching feature,the processing technology of sampling and weighting is used to predict the possible matching region of the face,and the maximum posteriori estimation is employed to achieve fast similarity matching.Experimental results show that the weighted histogram of the face in HSV has clustering characteristics, which can adapt to complex background interference.At the same time,the means of region prediction and maximize estimates of the SMCF is effective to reduce the computational overhead and to increase the reliability of the matching.The proposed method can be applied in face tracking,video surveillance and digital entertainment field etc.
出处
《智能计算机与应用》
2012年第6期28-30,共3页
Intelligent Computer and Applications
基金
佳木斯市重点科研项目(12002)
佳木斯大学科研项目(L2011-003)
关键词
视频流
人脸匹配
SMCF
加权直方图
Video Stream
Face Matching
Sequential Monte Carlo Filter
Weighted Histogram
