Ask Your Question

chris_more's profile - activity

2015-03-11 19:20:30 -0500 received badge  Student (source)
2015-01-21 15:21:07 -0500 commented question Camera Calibration with only one view. Why does it work ?

Spatial Augmented Reality (SAR). in my case a simple combination of a projector and a depth camera where I want to find out the intrinsic parameters of my projector (projector modelled as an inverse camera) and the extrinsic parameters to a given world system.

2015-01-21 10:16:20 -0500 asked a question Camera Calibration with only one view. Why does it work ?

Hi together

Background to my problem.

I am building an SAR application and using the calibrateCamera Methods to calibrate my projector and to my depth camera.

I can only use one planar surface to project my calibration pattern onto so I only can obtain 1 view for the calibration. Surprisingly the reprojection error is quite low and I get a good calibration. As far as I know one needs at least 2 different views of a planar pattern to calibrate a camera.( Zhangs approach)

Now I am asking myself why does this even work at all with only 1 view?

So does anybody know and can explain why it also works with only 1 view of a planar pattern?

Maybe is it because you only need at least 2 views to calculate an initial camera matrix but 1 view is enough for the Levenberg-Marquardt optimization algorithm, when using a good initial camera matrix?

Does anybody know the answer or has any hints