distortion correction using one image

asked 2014-09-09 08:12:27 -0500

M1900 gravatar image


I would like to calculate image distortion (radial and tangential) using only one image. I read the documentation of the "Camera Calibration and 3D Reconstruction" and recognized that theoretically at least two images are required and for better results more than 10 images are recommended.

In my application I do not need to calculate the orientation of the calibration target (chessboard) with respect to the camera since this is fixed. This means that the extrinsic camera parameters are given. The calibration target will always be parallel (with certain tolerance) to the image plane. Moreover, the skew coefficient can be set to zero and the focal length is known. I also guess that the principal point can be set to the center of the image.

My questions is if it is possible to use the functions "Camera Calibration and 3D Reconstruction" for my problem? Is it possible to reduce the degree of freedom for the calibration by simultaneously reducing the input data (one image only)? I could of course use multiple images that are identical except for the noise.

Thanks for any help!

edit retag flag offensive close merge delete