[ptx] flat mosaics

UteKarsten.Kiessling at t-online.de UteKarsten.Kiessling at t-online.de
Sat Apr 3 18:58:00 BST 2004


> >  Hello Pablo,
> >
> >I often need flat mosaics in exact parallel perspective. As you 
> >certainly know this is not possible in Panotools. Here you can only 
> >choose a very small angle of approximately 1 degree for the angle of 
> >view and then optimize either view, roll, d, e or view, yaw, pitch, roll 
> >for every dependent image. This is discribed by Helmut Dersch for a 
> >MF-Scan. But for long facades mosaics this isn't correct enough.
> 
> I've done this when making mosaics of multiple maps, i.e. really 
> pure/flat image stitching, with no projection and/or lens distortion.
> 
> It can work OK by locking (a,b,c) to zero, and fixing FOV for the first 
> image, then allow all other images modify only vertical and horizontal 
> shift, along with roll and HFOV (to comensate for verying image sizes).
> 
> When this works, it works very well indeed.
> 
> However, afaik, only PTAssembler supports directly these parameters, not 
> Hugin.
> 
> Terje Mathisen

Hello Terje,

thank you for your answer. But I see, that I have to explain the problem more clearly. What you suggest is nearly the same what I have wrote and done on small mosaics (BTW HUGIN supports these parameters to). But this is not parallel perspective. Perhaps this is more clearly on two extreme examples:

- For a long wall or a mosaic floor with 100 images, with every image HFOV = 1 degree, you get a result image with 100 degrees and you leave rectlinear projection and there is mutch projection distortion.

- For 400 images you get a result image with 400 degree and you leave cylindrical projection.

So what is needed for flat mosaics is the optimisation of:
 shift parameter d and e, 
 roll,
 a scale factor for every image,
 (and perhaps distortion parameter a, b, c and shearing parameter g and t to)
 
 Best regards
 
 Karsten


More information about the ptX mailing list