Disparity definition in LF matching

[MGMBarcelona]

Hi all,

Now I am reading the paper  Learning Sub-Pixel Disparity Distribution for Light Field Depth Estimation

So, I found an Eq. form wich follows 

  L(u, v, x + (u − uc) · d(x, y), y + (v − vc) · d(x, y)) - is it true?

Cause  for diagonal shifted optical centers   this Eq leads to biger optical base and the saim disparity we should obtain the saime depth. That contradict to the triangulation principle. 

Maybe this fact comes from the Fourier lens setup, which is not invariant to rotation?

Thanks for repaly.

Mikhail