One of the key issues of Augmented Reality (AR) is the camera calibration. In AR applications, camera calibration has some special requirements; it prefers good registration of virtual objects in real environment to high precision. We propose a flexible and robust technique to calibrate the camera, it is well suited for augmented reality applications. This technique only requires the camera to observe a planar pattern, e.g. a normal wall. The proposed procedure consists of a closed-form solution, followed by a nonlinear refinement based on the maximum likelihood criterion. We have tested our proposed technique in the real environment, and have obtained very good results. Detailed analysis is also included in this paper. Compared with classical techniques which use expensive equipment such as two or three orthogonal planes, our technique is easy to use and flexible.