Assuming that I have the location and rotation angle of the camera, is there a method in projection geometry to determine the skew of the image? I want to correct the camera position to have unskewed image by computing correct location and rotation angle of the camera. 
Asked
Active
Viewed 430 times
1
user0193
- 313
-
1Unskewed image you can obtain if you would have optical axis of the camera perpendicular to the marker plane and crossing the marker in the middle. – Widawensen Jun 15 '21 at 10:01
-
@Widawensen yes, but remember its not aways gonna be a plane, and its hard to determine the normals to the surfaces and objects to align the optical axis of the camera. I am looking for something more deterministic then iterative as i have a planning problem to solve. – user0193 Jun 15 '21 at 11:27
-
1If the marker is a square on a plane with known dimension it is possible to determine it's location in the camera frame, do you use this possibility? – Widawensen Jun 15 '21 at 11:39
-
@Widawensen sure, thats calibration. I understood what you are saying. – user0193 Jun 15 '21 at 15:04
-
1Anyway, see for example https://robolab.sjtu.edu.cn/kindeditor/Upload/file/20170830/20170830144744_8958.pdf for details – Widawensen Jun 15 '21 at 15:12
-
@Widawensen thanks. Pose estimation is easy SIFT, RANSAC. My concern is to correct physical camera so that the image is close to isomorphic, to make such correction, the "Degree of perspective is to be determined" which you already gave me a hint i.e surface normals. – user0193 Jun 15 '21 at 15:25