Hi Rick,

On December 27, 2010 06:20:34 am RueiKe wrote:
> Thanks Yuv for your very detailed response!

happy to read that it helped.

 
> I can usually acheive about 0.3 pixel mean error for a 16k wide pano.

beware of this metrics.  It is blind to the quality of the CPs and thus to the 
real quality of the panorama, and I think that if we want to make Hugin more 
user friendy, we should remove the number all together and replace them with
"likely to be very good / well / poorly / badly aligned" instead.

I've had a long discussion about this metrics with somebody else on this list 
a few months back.


> For this project, the intital alignment looked fine.  It was when I
> removed the outlier control points and realigned when I ran into
> problems.

removing the outlier is a blind process as well.  If you have, for example, a 
lot of features in the foreground that "captures" the majority of the CPs, the 
background CPs will be declared outliers, even if it is known that the further 
away from the camera a CP is located, the less sensitive it is to parallax and 
other negative factors.  And if those front features are in movement, the 
statistics are toast. 


> Seems like my choice for optimization parameters may have
> caused the problems.

probably the immediate one, yes.  It seems to me that you have the potential 
to bring your process to the next level and if you want you will find helpful 
feedback here on the list.


> Let me try it again when I finish my current
> project.  Should be able to report back in a few days.  If I continue
> to have problems, I will open a bug report. 

OK.  Looking forward for your report - preferably a link to your flickr 
gallery ;-)

Yuv

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to