I'm trying to figure out how to detect/compute the camera's focus of expansion coordinates before computing the camera translational and rotational parameters in order to improve the camera motion estimation. I use matched features between frames to compute the optical flow field, and having the focus of expansion coordinates will improve the estimation of the camera roation and translation.
I've been following the paper by irani for egomotion recovery and couldnt really understand how they do that coordinate extraction. (link below)
If anyone can give me an insight on how to retrieve such coordinates or on how to estimate the camera's motion parameters in a different way i'd really appreciate it.
Thank you in advance
http://www.vision.huji.ac.il/papers/ego-mtn-cvpr94.pdf)