The invention discloses an inter-frame information oriented head gesture estimation optimization method. The method comprises three steps of face detection, face alignment and parameter estimation. During face detection, according to the face position information of a previous adjacent frame, an area to be detected is reduced, and face detection speed is improved; during face alignment, accordingto whether the head gesture of the previous adjacent frame is subjected to overlarge deflection or not, feature points which are positioned on the edge of a face contour to cause inaccurate positioning are removed so as to improve robustness; and during parameter estimation, the similarity of the parameter in previous and rear frames is used for simplifying a solving process, and a direct linear transformation step is omitted. By use of the method disclosed by the invention, on a premise that a complex operation is not increased, the accuracy and the instantaneity of a traditional head gestureestimation method are optimized, and a support is provided for the method to be combined with other relevant technologies to realize rich application.