I have implemented and tested in python the recoverpose () for standard images and it seems to work very well.
If I try to use it for panoramic images, in particular for equirectangular images, the angle in degrees of euler, or more generally the rotation matrix, is always equal to zero or to a very small value even if its images have a rotation evident among them.
If I draw the epilines:
for standard photos they are slanted and have a different slant in the two images.
For equirectangular photos, they are all horizontal and identical for both photos.
The two panoramic photos were taken a few meters away from each other, and considering the center of the photo as a reference angle, the two photos have a 90-degree rotation.
For equirectangular panoramic images, do I have to use other functions, other flags?
Where am I wrong?
Thanks so much