As a ball has the advantage of contour continuity, it has been used widely in camera calibration, especially in multiple camera calibration. Using a ball as calibration target can make up for the shortages of plane target that it may distort or even disappear in the field of view when the viewpoint is large. However, when a ball is projected into the camera it is usually not a standard circle, but an ellipse. The calibration precision will be affected when the elliptic geometric center is not the same as the ball-center′s real image. There are two factors that determine the distortion: the relative size and position of the ball with the camera. The influence of the two factors are analyzed, and the distortion rules are studied by simulating ball′s imaging model. A correction model which is verified feasible and effective in experiment is established. Through the model, the ball center′s projection error can achieve subpixel precision.