[ptx] Re: Hugin stitching not working

[Sean Davis]

Ok...

This is really weird.  It looks as though the problem is due to the fact that 
I needed to copy my .pto file to /tmp/PT_stitcher.txt for some reason.  When 
I did that, everything worked fine with the stitching.  Without doing that, 
the optimization worked okay, but the stitching was not reading the 
parameters in from the optimization.  Very weird. ...Well, everything works 
okay now, with the quirk that I have to place the .pto file in /tmp to get it 
to work.

-Sean

On Thursday 07 July 2005 13:55, Carl von Einem wrote:
> I guess that's better be asked to the list. I think the Commands for
> Panorama Creation are strange: do we still need a -buf / +buf? I thought
> this was something needed only in older versions of PTStitcher for
> certain file formats (TIFF with masks?)...
>
> when I import your values in my corresponding PTMac project file I see
> that v64 (HFoV) in your p line and also the width / height values are
> not what PTMac would consider. I'd also prefer a different output file
> format.
>
> I did some more optimizing and arrived at
>
> # script file for ptStitcher created by PTMac
>
> p w2320 h2126 f0 v95 u0 n"TIFF_m"
> # line beginning with p defines the output image
> # I used a different interpolator and file format
> # also the size of the resulting pano is different, see?
> m g1 i2 f40
> # who can explain the m line? It's for enblend, isn't it?
> i n"045.jpg"
> o f0 y0 r-1.33247 p8.54754 v39.1991 a0.002534 b-0.010434 c-0.008989
> d-8.76285 e11.0808 g-2.87594 t-0.090636
> # o: output line for the stitcher
> i n"046.jpg"
> o f0 y23.1143 r0.400612 p10.0231 v39.1991 a0.002534 b-0.010434
> c-0.008989 d-8.91023 e11.7722 g-2.87594 t-0.090636
>
> Here are my results for the Grand Prix (European Song Contest joke)
> 12 points to Germany ;-)
>
> see my attachment. Does your project stich with this script? I
> successfully stitched them but the resulting TIFFs are way too big. I'll
> send you a JPG of the resulting pano directly.
>
> Carl
>
> Sean Davis wrote:
> > Hi Carl,
> >
> > Thanks for helping me on this.  I have attached the new test.pto.  It
> > looks like I am getting ypr values similar to yours.  However, now when I
> > look at the preview I can only see one image at a time!  If I try to
> > stitch the images, I just get one image back!  what am I doing wrong?
> >
> > -Sean
> >
> > On Thursday 07 July 2005 12:20, Carl von Einem wrote:
> >>Hi Sean,
> >>
> >>reading those files was a hard task as a Mac user, believe me! In the
> >>early days of the panotools that was the only way: learn to read and
> >>write the script files for PTOptimizer and PTStitcher manually. Today it
> >>sometimes helps when you can interpret them... :-)
> >>
> >>A good starting point for tutorials is the wiki at panotools.info
> >>
> >>The new .pto looks weird: nice Control Point Distances, but strange
> >>results for at least y (yaw). I see you had a preset yaw of -7.39216 for
> >>the second picture (No. 1) before optimization.
> >>
> >>Can you try to reset all your values to 0 again and than do a new
> >>optimization? Choose all control points to be used for optimization, and
> >>then step by step
> >>- only optimize yaw of one picture
> >>- new optimization for pitch and roll of both pictures
> >>
> >>Have another look at my values for your images, you may also cheat and
> >>use those as starting points (insert those values in hugins tab for the
> >>image parameters):
> >>
> >>image "045.jpg"
> >>yaw: 0
> >>roll: -1.36353
> >>pitch: 8.94349
> >>
> >>image "046.jpg"
> >>yaw 22.2485
> >>roll 0.42884
> >>pitch 10.4035
> >>
> >>Note: it's always easier for beginners to stitch a full cylindrical
> >>pano, set enough control points (up to 7 per image pair are enough),
> >>never forget the last image pair (last image and image0), anchor yaw of
> >>one image, optimize y p r for the rest and fov. A wavy horizon can
> >>easily be prevented by setting t1 control points as I described earlier.
> >>BTW a wavy horizon can also be corrected in a second run.
> >>
> >>happy stitching,
> >>Carl