An improved continuously adaptive MeanShift (CamShift) tracking algorithm,which uses the hue and differential information to build the target model,is proposed to improve the performance of the conventional method. The maximum of the difference values of the eight neighborhoods is used as differential value,which describes the detail information and the relative position information.Based on the hue and differential joint histogram of the target model,the hue and differential joint probability distribution image of the tracking image,which c an reduce the disturbance of the redundant information generated by independent modeling for different characteri stics,can be got by the back projection.The target positioning can be completed by MeanShift method in the tracking window.Furthermore, the adjustable size of the target is limited to prevent the disturbance of backg round information.Simulation results show that the improved CamShift tracking method has good tracking performance.E specially,the valid target positioning and tracking can also be completed when the background or some areas in the background are similar to the target,which can enlarge the application scope of the CamShift tracking algorithm.The computation time of single frame image is less than 30ms,which can meet the real-time request of the tracking system.