In gray-level image sequence,due to the sensitivity to illumination and the lack of the information for the object representation,the face tracking based on the mean shift algorithm may be unstable.In order to apply mean shift algorithm in gray-level image effectively,a strategy which integrates intensity and orientation codes to represent target is adopted.A new algorithm is presented for face tracking in this paper.Experiment results show that the new algorithm can successfully cope with clutter,partial occlusions,illumination change and target variations such as scale and rotation.The computational complexity is greatly reduced.If the size of the target is 48×68 pixels,the computation time of the whole algorithm is about 34ms.