The invention discloses a method for obtaining a clear face image of a moving human body in a monitored video. Firstly, after a moving human body is detected, an AdaBoost learning algorithm is adopted for detecting the direct human face in a rectangular region which is 1 / 6 of the length of the upper body part; secondly, whether the face area is larger than 10 percent of the whole image area S is judged, if the face area is less than 10 percent of the whole image area S, the face is tracked, and optical parameters of a camera are controlled so that the face area can reach 10 percent of S; if the face area is not less than 10 percent of the whole image area S, whether a definition evaluation function takes a peak value is judged, if the definition evaluation function takes the peak value, the face image is clearest; and finally, the clear face image is saved, and the optical parameters of the camera are controlled so that the camera can be restored to a normal field monitoring status. The invention realizes feature emphasis of the face, is easy to carry out and can obtain the clear face image from the monitored video with low resolution in the case of long distance.