Abstract: Under the framework of classical Mean Shift tracking algorithm,color is an effective vision characteristic.But in practice,the validity of color will be seriously weakened in practice because of the variety of the illumination,the angle and the location of the vidicon,Consequently,the tracking will be instability.In this paper,a robust Mean Shift face tracking algorithm under varying illumination is proposed.First,modeling and analyzing the effect of illumination.then according to the results of the analysis,bring the truncated Gaussian kernel function into the probability density function to reduce the impact of the illumination.At last,bring the new probability density function into the framework of the Mean Shift tracking algorithm.And the result of the emluator indicated that it can achieve a robust face tracking under varying illumination conditions.