In this paper we present a camera calibration method using Snell’s Law. Traditional camera calibration is based on the pinhole model, which is an approximation algorithm using untrue geometrical assumptions and giving a single lumped result for the various optical elements in the camera system. Using full modeling of lens geometry, the proposed method establishes the geometric relationship between images and objects via Snell’s Law. A matrix equation that relates the intrinsic/extrinsic parameters of image the plane and six pose parameters of each element is determined from sensitivity analysis. These parameters can be identified using the least square method by observing points with known coordinates. An illustrative example using a two-camera stereo coordinate measurement system demonstrates that system performance via the proposed method is better than the pinhole model.