[ptx] Q: Possible bug in Autopano

[alexandre jenny]

Hi,

Yes, it seems definitely to be a bug in the last stage of the sift keypoint
calculation. I will have a look on it.
Note : this bug should not be influenced by the command line parameters.
Do you see that ?

Alexandre Jenny
http://www.kolor.com
http://autopano.kolor.com


> -----Message d'origine-----
> De : ptx-bounces at email-lists.org 
> [mailto:ptx-bounces at email-lists.org] De la part de Morten A. Steien
> Envoyé : mercredi 14 avril 2004 12:21
> À : ptx at email-lists.org
> Objet : [ptx] Q: Possible bug in Autopano
> 
> 
> Before I start complaining let me just say that Autopano is 
> supercool and is a great help when creating panoramas.
> 
> I have used Autopano 1.02 quite a lot lately and I have 
> repeatedly come 
> across
> a problem. Some control points looks to be the correct object but the 
> actual
> mark is placed differently in the two pictures. It looks like 
> the object has been mirrored or rotated before selecting the 
> actual XY value. The 
> difference
> looks to be 20 to 25 pixels which looks much like sqrt(512)
> 
> See the example with control point nr. 15 here:
>   http://home.powertech.no/morteas/temp/1.jpg
>   http://home.powertech.no/morteas/temp/2.jpg
> 
> I use autopano with the following parameters:
>   autopano.exe /search:1 /ransac:1 /keys:20 /nomove /size:512 
> /ptgui <dir>
> 
> There are (relatively) few real false control points so if 
> these that are nearly correct was fixed it would shorten the 
> workload significantly when making panoramas.
> 
> Has anybody else come across this problem or even better, a solution?
> 
> -- 
> Best Regards,
> Morten A. Steien
>