[ptx] a few bugs

[JD Smith]

> With optimization it appears, that program poorly handles the bigger numbers 
> of images.
> I had problems with optimization of rows of 15 pictures. The optimization 
> results displayed are mach to big.
> When stitching a couple of pictures only usually results with the following 
> numbers (optimizing position and barrel distortion):
> average c.p. distance:   0.199933
> standard deviation:       0.234668
> maximum:                    0.655413
> 
> with 15 pictures numbers were:
> average c.p. distance:   25.560242
> standard deviation:       52.280548
> maximum:                  336.597000

Just a comment that the numbers here are somewhat meaningless, since
they depend on the target output size of the panorama.  The bigger the
size, the bigger the numbers, even for the exact same optimization
quality.  I usually have Hugin estimate the size quite frequently to
make sure that they remain consistent, but occasionally the size gets
much too large, causing the errors to appear larger than realistic.  

I think quoting offsets not in pixels in the final panorama coordinate
system, but in seconds or minutes of arc would be far more useful, or
even as a fraction of the HFOV of the source images.  Providing pixel
unit quotes in the final panorama space is not a bad idea, since it
tells you how bad mismatches will affect the final product, but for the
purpose of refining the optimization, and providing an *absolute*
quantity for discussing the quality of a given optimization (e.g. on
this group), a more output-agnostic measure would be more appropriate.

JD