Calibration

Distortion: 
    The figure shows an image of a grid taken by one of our cameras. Important radial distortions are present.
    These distortions are not taken into account in the pinhole model. To deal with them, first order Seidel distortions are introduced.

    If x and y are the image coordinates without distortion and x' and y' the coordinates with distortion, the Seidel model states:

    x' = x + d R (x-Px)
    y' = y + d R (y-Py)

    where d is the intrinsic parameter of distortion and

    R = (x-Px)2/dx2 + (y-Py)2/dy2

    Once "strong" calibration is performed, it is possible to foresee where the image of a point in a scene will be on the image. To see that, coordinates of a grid caracteristic points have been measured. An image of this grid is taken. The position of these points on the image is foreseen through "strong" calibration parameters. The following figure shows these previsions surimposed on the original image. These projections have been estimated by the pinhole model with distortion.