[ptx] hugin, auto selection of optimization values

[Pablo d'Angelo]

Hi Bruno,

Bruno Postle wrote:
> This is a bug that appeared a few months ago, previously hugin did the 
> right thing, I'll try and describe it.
> 
> It's easier to see if you load a single image, though the problem is the 
> same with multiple images.
> 
> Select a pair of vertical or horizontal control points, and optimisation 
> for 'roll' is automatically selected, OK.
> 
> Select two pairs of vertical control points and optimisation for both 
> 'roll' and 'pitch' is automatically selected, OK.
> 
> Select two pairs of horizontal control points and optimisation for 
> 'roll' and 'pitch' is automatically selected.
> 
> I can't think of any circumstances where horizontal control points could 
> control 'pitch' in any useful way, whereas they would control 'yaw' when 
> using rectilinear output.

Thats true. For rectilinear output its probably a good idea to do that.

But for equirectangular, I think both values can (and need) to be 
optimized. I used horizontal point only on the horizon for this. Imagine 
one of these slightly bending "snake" panoramas. Assume the "anchor 
image" is on the upper bend of the selected images. Now pitch needs to 
be adjusted for that image as well, to move the horizontal line into the 
horizon.

Just try with a 2x6 mosaic or similar, and select 2 horizontal control 
points, and an anchor image that is below the horizon. the optimizer 
will figure out the correct pitch for the reference image. I think this 
was what I thought when adding that...

unfortunately I forgot the rectilinear case, I will add your patch only 
for the rectilinear case.

> Just to prove I'm not completely useless, here is a patch that fixes 
> this behaviour.

On the contrary, you have always been great, doing many important bigger 
and smaller things in the background.

ciao
   Pablo