一种改进的窄带算法在彩色序列图像人脸跟踪中的应用

该文提出了一种基于几何主动轮廓模型的人脸跟踪方法.通过直方图反向投影,使人脸区域表现为一个一致性区域与背景相区别.研究了一种改进的窄带算法实现曲线演化:以等间隔分布的节点表示运动曲线,只在这些节点上计算Level set函数的变化值,窄带区内其余点的Level set值的更新通过插值和查表的方法实现;根据节点的局部图像信息决定节点的运动方向和时间步长值.实验表明该算法能在满足一定精度的前提下,快速地对运动人脸进行跟踪.

Tracking of the Facial Region in Color Video Sequences Using an Improved Narrow Band Algorithm

A novel method based on geometric active contour model for face tracking is presented. After performing the chromatic histogram backprojection, facial region appears as a homogeneous region in the backprojected image. An improved narrow band algorithm is developed for implementing the curve evolution. The propagating curve is represented as a set of nodes. The update of the level set function is only calculated on these nodes, while the evolution on other grid points is performed using the interpolation and lookup table method. The motion direction as well as the time step of each node is different with respect to the local image properties. The experimental results indicate the proposed algorithm is fast and accurate enough for tracking faces in color image sequences.