【作者】 胡韦伟； 汪荣贵； 胡琼； 李守毅；

【Author】 HU Wei-wei WANG Rong-gui HU qiong LI Shou-yi Faculty of Computer and Information, Hefei University of Technology, Hefei ,230009, China

【机构】 合肥工业大学计算机与信息学院；

【摘要】 提出一种在视频中跟踪人脸的新方法。该方法首先用肤色特征检测出第一帧中的肤色区域,再计算出人脸的大小和位置,用来初始化卡尔曼滤波器,同时对满足条件的肤色像素进行高斯建模。然后用卡尔曼滤波器预测出下一帧人脸的大小及位置,再用高斯肤色模型检测人脸,最后根据新检测到的像素更新肤色模型。实验结果表明,该方法可以实时准确的跟踪到人脸,并对不同光照,背景,角度的场合具有一定的鲁棒性。更多还原

【Abstract】 This paper presents a new face tracking method in video. Firstly, detect the skin-color region from the first frame using skin-color feature and compute the position and size of the face which used to initial the kalman filter. Then estimate the size and position of the face in the next frame using the kalman filter and model the Gaussian model using the detected skin-pixel. Finally use the Gaussian skin-color model to detect the face and renew the model. Experimental result shows that the human face are exactly tracked in real time, and the algorithm is robust to different illumination ,background ,and angle of view.更多还原