A critique of the ptgui interface

[Bruno Postle]

On Wed 17-Apr-2002, Kai-Uwe.Behrmann@VHBAC.SMF.SACHSEN.DE wrote:
> 
> Let us take an suitable photo. Most useable is someting from the
> artificial world - architecture, a room or a testgrafic with many
> lines on it. This foto should show after lenscorrection all lines
> linear and not curved ( barrel-distortion ).
> 
> As well the lines or edges should show no color-displacement. And
> there should be no darker edges visible like with many lenses if taken
> the foto with wide opend lens. 

That is the problem.  Ptoptimizer and ptstitcher don't suppport
different a,b & c values for each colour like the plugin.  Ptstitcher
doesn't support radial brightness correction either.  The plugin doesn't
suppport d & e parameters, so some people have to pre-process their
images in the plugin before stitching them.

I don't think that there is any way of doing this in one step with
Helmut's tools.

> With an softwareline like the distance/arc-messuringtool in gimp it
> should be easy to detect barrel distortion. Simply if one can change
> the related values in the lensgui and see instantly what happens, the
> good values are more easy to been find --> continuously showing the
> new image over the old one to compare both. 

This is how I calculated 'b' for my old camera.  I took a picture of a
grid and tried lots of values in the correct-plugin until I was happy.

However, it's _much_ easier to do this with the ptgui/ptoptimizer
combination.  Just take 5 or 6 photos in roughly the same direction
(this doesn't need to be a 360 panorama if you already know the hfov of
your lens).  Then select _lots_ of control points and just optimize
roll, pitch, yaw, a, b, c.

Ptoptimizer will calculate a,b,c,d & e for you, very accurately.

> At the moment its an timeeating try/error testing with editing/saving
> the script and starting the optimizer and ptstitch by hand.

It is _painful_ doing it by editing the script files, that's the process
that ptgui manages.

-- 
Bruno