It has been difficult problem in face tracking and recognition that of the coveraged target face,wide-angle tilt and rotation can cause tracking lost.For above issues,a time rewinding face tracking algorithm based on Haar feature detection and improved Camshift is proposed.Time variable is introduced in Haar classifiers to realize time backtracking and the change-weighted histogram is added to the probability density image to increase the weight of face region.Supplemented with light compensation and ellipse template matching,the new algorithm effectively overcomes the problem of face coverage and target lost,and is more effective than other algorithms.