一种基于空间直方图的Mean-Shift跟踪算法*

经典的Mean-Shift跟踪算法以颜色直方图为特征对目标进行跟踪。颜色直方图反映的是图像中颜色的组成情况，即出现了哪些颜色以及各种颜色出现的次数。颜色直方图具有旋转不变性、缩放不变性等优点,经常用于图像检索，即通过比较颜色直方图的差异来衡量两幅图像在颜色全局分布上的差异。但是颜色直方图不能反映颜色的空间分布特征，当跟踪目标与背景色颜色相近时可能造成错误跟踪，导致跟踪失败。考虑目标颜色空间分布特征,将空间分布信息融入颜色直方图中，提出了基于空间颜色直方图的Mean-Shift跟踪方法，全面地反映了颜色的整体分布信息和空间分布信息。在VC 6.0上利用新方法和经典Mean-Shift跟踪方法分别对发生遮挡的视频和快速运动的视频进行跟踪，实验结果表明，提出的新方法能够更好地对目标进行跟踪，避免跟踪过程中的目标丢失。

Improved Mean-Shift tacking algorithm based on spatial histogram

The color-histogram indicates the color components of the target object-the sorts of color and the number of every kind of color. The global color histogram method has some excellencies-the rotation inflexibility and scale inflexibility and so on, so it is often used in the image retrieving-comparing two images through computing the difference of the color histogram of the two images. But the color histogram does not involve the spatial distributing information of the color .It may bring on failed tracking while the colors of target object is similar with the background colors.This paper proposed an improved tracking mehod which based on spatial color histogram, which involved both the colors number information and the spatial distribution information of the colors information. This paper applied new method and the classic Mean-Shift tracking method on the object occlusion of video and fast-motion video using VC 6.0. The experiment shows that the new method can tacking the target object better than the classic Mean-Shift tracking algorithm.