Experiment on Parameter Selection of Image Distortion Model

Most of current camera calibration methods for a non-metric digital camera adopt polynomials of image coordinates composed of terms representing the correction to the principal distance, the offsets of the principal point, the radial lens distortion, and the decentering lens distortion of the target camera as the image distortion model. However, there is no standard procedure to evaluate appropriateness of parameter selection of the image distortion model. Therefore, we conducted a field experiment on parameter selection of the ordinary image distortion model widely used for camera calibration. We adopted a calibration method using a set of calibration points distributed on the 2-D plane with no ground survey. Four non-metric digital cameras were calibrated in the filed experiment. 16 rounds of camera calibration for six different parameter sets of the image distortion model were conducted. Evaluation of calibration results was performed by differences of image distortions calculated at all pixels on image between the obtained image distortion models. The experiment results indicate that the adoption of the decentering lens distortion component has the significant influence on the estimation of the distribution of image distortion, even if the magnitude of the decentering lens distortion component is small.