camera intrinsics calibration produces wild out-of-range principalPoint

Hi -

A student and I are performing calibration with estimateCameraParameters and checkerboard patterns. The error in reconstruction is low, but the principal Point returned is outside the range of the pixel array of the camera. For one of our cameras, it is negative when we fit with EstimateSkew=true (which causes an error when we try to create a camera object). But even if we set EstimateTangentialDistortion=false,EstimateSkew=false the principal point is out of range. The detected grid locations seem fine (they are on the image).

Any suggestions?

The calibration grid is sometimes a little blurry because it is small and the lens doesn't have a large depth of field (so sometimes the edges are just a bit blurry).

Thanks

Steve

[Giridharan Kumaravelu]

We would need more information to reproduce and investigate this issue that you are facing.

Could you provide your calibration images and the version of MATLAB you are using?

[Stefano]

I am here a while later with a probably similar error. I work with a small calibration pattern as you do (square size 3.75mm). The new version of Matlab (2025a) pop an error that claims patterns may be too similar across images. By doing debugging over matlab functions (modifying matlab functions to avoid sanity check and obtain the camera parameter object), I found out that the principal points is very off from the center, and occasionally negative - which indeed gives an error when, e.g., calling params.Intrinsics (see attached matlab output).

I still did not find a way to fix this. Did you manage to fix this issue somehow?

Error using vision.internal.cameraIntrinsicsImpl>parseInputsSimulation (line 240)

The value of 'principalPoint' is invalid. Expected principalPoint to be positive.

Error in

vision.internal.cameraIntrinsicsImpl>parseInputs (line 213)

r = parseInputsSimulation(defaultParams,varargin{:});

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in

vision.internal.cameraIntrinsicsImpl (line 82)

r = parseInputs(varargin{:});

^^^^^^^^^^^^^^^^^^^^^^^^

Error in

cameraIntrinsics (line 17)

this = [e-mail](varargin{:});

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in

vision.internal.calibration.CameraParametersImpl/get.Intrinsics (line 504)

intrinsics = cameraIntrinsics(this.FocalLength, ...

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Error in

main00_calibrationSingleCamera (line 106)

K = params.Intrinsics.K;

^^^^^^^^^^^^^^^^^^^

[Steve Van Hooser]

Hi Stefano - Sorry, I don't have any updates here. Good luck! Best, Steve

Answer 1

[Matt J]

You haven't provided data with which to repeat the calibration, so one can only guess. But if reconstruction error is low, the principal point is probably correct. It might mean you have a weird camera, but there is no fundamental requirement that the principal point lie within the pixel array boundaries.