Calculating line of sight angles from two different photographs

[drewandsarah1]

I have 2 photographs of the same side of a randomly textured sphere taken simultaneously from different angles. How can I calculate the angle between each photographers line of sight?

The only available data is the texture on the sphere; distance between photographers, etc, etc are not available.

The findings are part of a larger scientific project which will be published so I need to achieve a very high level of precision.

Thank you to whoever can help.

 

[Paul Belino]

You would really need to calibrate each camera first by photographing a flat grid of points (X,Y) at a known distance Z face on to each camera. Hopefully where the sphere will be.
This will enable distortions (optical aberrations, focus errors, magnifcation errors, keystoning etc) and angle of views to be assigned to an individual camera.
No two lenses can be the same or even be set the same.
From the known (X,Y,Z) grid coordinates so this will enable each camera's pixel position to be assigned a (Theta,Phi) altazimuth and altitude angle on a digital image from each camera at its own Z.
Getting to the sphere....
You need reference points on the sphere that are common to both cameras. The more and the more spread out have the more accuracy.
So you know the (X,Y,Z) position of each camera .
You know the (i,j) pixels position of each reference point in each camera.
You know that a line from those pixels, from each camera, must intersect in space at the reference points.
This will give you the (X,Y,Z) of each reference point.
This will enable to overlay the two digital images from each camera accurately
You now have enough information to obtain the (X,Y,Z) points of any other feature on the sphere common to both cameras.

 

[drewandsarah1]

What I ended up doing was plotting the center point of each sphere as if it were a circle, then moving each common centerpoint to a graphical representation of the same sphere with GPS like overlays on it. I then converted the distance between each centerpoint to degrees.

Is there a reason that would not work? I only need to be accurate to a full degree.