基于灰度分布的概率密度实现实时图像分割

针对实时图像质心跟踪的问题，提出了一种目标分割的算法。对相邻两帧图像的跟踪框和背景框的灰度分布统计，加入滤波因子，再根据Bayes判别规则构造判别函数，把背景像素和目标像素进行分类。实验结果证明，该方法克服了由于图像序列中某一帧图像灰度分布发生突变产生的影响，能较好的分割出图像中的目标。在电视跟踪系统中有效的保证了跟踪的稳定性。

Probability density of gray scale for segmenting real-time images

In order to realize centroid tracking in real-time images,a method of segmentation is proposed.It combines the statistic of the gray scale in track and background gate of two contiguous frames with a smoothing value.Furthermore,according to the classification criterion built by Byesian decision theory,the background and the object are classified.The experiment results show that this proposed method can avoid the effect that attributes to the wave of gray scale in some frames of an image sequence,and can segment the object from an image better.It can ensure the stability of tracking effectively in the video tracking system.