Determine the Center of Rotation from a set of images

I would like an efficient algorithm to determine the center of rotation from a set of N images. The set of images are of the same scene in which the camera was rotated by a mechanical axis. For sure, the center of rotation is within the image, very close to the center of the image. I do know the amount of angular rotation of scene for each image, as this was commanded by the robot.

I can envision a brute force method using image registration with 3 dof (x,y,theta). However, I feel that there exists a more elegant way to get this done using a transform or some other method like taking the average of all images and looking for the "North Star" point. Thanks.

[Morgan Facchin]

Hello,

This post is a bit old but I'm asking anyway, I am trying so do the exact same thing, did you find any nice way of doing it? I also had that "North Star" thought but still can't do anything satisfactory!

Thanks ;)

Answer 1

[Image Analyst]

Maybe use optical flow? Not sure it's any better than the method you're already using though.

[James Eder]

Thank you for the suggestion. It certainly points me in a direction I had not considered.