This paper presents the Bayes filter method to estimate the geometric parameters of a four-bar linkage. The input data obtained by tracking are the time history of the coordinates of some point rigidly attached to the links of the mechanism. This tracking data is corrupted by noise. The Bayes kinematic identification is designed and performed independently for each element of the linkage. The estimation of the geometric parameters is performed using constraints equations for each joint. The main advantages of the proposed Bayes kinematic identification strategy are: reduction of the geometric identification computational costs, numerical efficiency of the identification algorithm and improvement of the geometric identification results. Accurate identification of the parameters is important because the positioning accuracy of different mechanisms depends upon the kinematic model which describes also the geometry in a parametric form. The method presented is suitable to extract the planar geometrical models for mechanisms with revolute joints.