search

kocurvik / robust_self_calibration

31 stars 7 forks source link

Does the lens distortion affect the result? #5

Open huntkao opened 2 months ago

huntkao commented 2 months ago

AFAIK, lens distortion model matters a lot to the computation of the intrinsics, ie., focal length and principle points. Is there assumption that they are already known in this paper? Thanks.

kocurvik commented 2 months ago

Our method assumes that there is at most some negligible distortion.

Recently, we have had a paper accepted to an ECCV workshop titled "Are Minimal Radial Distortion Solvers Necessary?". Where we have evaluated some methods for radial distortion estimation using the division model. The methods output the radial distortion parameters of the division model + the fundamental matrix. The conference should start in a few days so the paper should be available soon.

You can check out the Python evaluation code in my rd repo. The solvers were implemented into Poselib and you can find them in the rd branch of my fork. I have also prepared a pull request to Poselib with methods for one of the problems. I will probably add the solvers for the other problem soon.