Fisheye distorted image calibration using parabolic perspective projection constraint

Fisheye lens is a kind of special lens that has larger field of view than the conventional camera, thus fisheye lens is widely used in many situations like security automation, robotic vision etc. But the image from the fisheye lens is also severely distorted. So correction is needed to restore image. This paper presents a simplified algorithm based on parabolic perspective projection, which is a precise approach of the projection model of a fisheye lens. We use checkerboard to calculate node pixels´ distance to the center of both corrected image and distorted image, and use these data to calculate the parameters of the parabolic perspective projection, and through the model we may generate a mapping matrix between distorted image and corrected image, and lens can be calibrated through the mapping matrix.