To exploit the accuracy potential of a video camera, all systematic errors must be modeled and compensated for. In this article, a new geometric camera model, where difference in scale and lack of orthogonality between the image axes is compensated for after, and separately from lens distortion, is proposed. The new model can be calibrated from both planar and non-planar calibration objects. The feasibility of the model is demonstrated in a typical camera calibration experiment, which indicates that the new model is more accurate than the traditional one. It also gives a simple solution to the problem of computing undistorted image coordinates from distorted ones. Further, the article suggests how to get initial estimates for all the camera model parameters, how to select the number of parameters modeling lens distortion, and how to reduce the dimension of the search space in the nonlinear optimization. There is also a discussion on the use of analytical partial derivatives.