彩色图像的实时人脸跟踪方法

本文提出一种彩色图像序列中的人脸跟踪方法。本文所提出的跟踪模式包括失脸肤色模型及运动模型。在人脸肤色模型中，首先将图像的ＲＧＢ空间变换到色度空间，然后利用最大能量坐标和矩表征色度空间的直方图分布。运动模型由运动检测与运动预测构成，以减少搜索区域，提高检测速度。最后，介绍根据上述技术实现的在线人脸跟踪系统。实验结果表明，本文提出的跟踪方法可以在复杂背景中实时地跟踪人脸的自由运动。

REAL-TIME HUMAN FACE TRACKING IN COLOR IMAGES

This paper presents a method of real-time human face tracking in color images. The tracking model contains Human Face Color Model (HFCM) and Motion Model. In HFCM, it first transforms images from RGB Space onto Chroma Space, then adopts quadrature of histogram to characterize human face. Motion model consists of motion detection and motion prediction in order to decrease searching region and raise tracking speed. Finally, an on-line system which is realized using the above two models is introduced. Experimental results show that in complex backgrounds, the tracking model presented in this paper could track human face's free movement in real-time.