[ptx] none ignores d, e parameters

Daniel M. German dmgerman at uvic.ca
Thu Feb 23 23:02:45 GMT 2006


 >> 
 >> I noticed the same behaviour I was debugging PTmender.

 Pablo> Did you notice any other different behaviour. I hope that was
 Pablo> the only one, except the different behaviour when cropping
 Pablo> circular fisheye images.

Hi Pablo,

I have been very busy in the last 3 weeks, and I'll probably won't
have much time for the next 4 weeks either, but after that one of my
plans is to write some test cases to compare both programs. I think it
will help both projects.

Does nona do feathering? If it does, the masks calculation and the
feathering will probably be different. I can explain you the masking
calculation algorithm used by panotools if you are interested (I don't
think it is very good, by the way). 

I have not implemented JPEG flattening. Once it is done the comparison
of both programs will be simplified.

With respect to fisheye images, I made the decision to not spend time
implementing it. I don't have any of these lenses, so the feature was
not in my priority list. Perhaps somebody else can take the task?

 Pablo> Now the only remapping difference between PTStitcher/PTmender
 Pablo> and nona is when using the S parameter in the script files to
 Pablo> describe cropping of circular fisheye images.

 Pablo> I believe this is a bug in pano12.dll.

 Pablo> the PTStitcher documentation explains that S is used to select
 Pablo> an area in the input image, while C crops the input image, and
 Pablo> applies all parameters (HFOV, etc) to the cropped area.

 Pablo> However, for almost circular fisheye images (8 mm fisheye on
 Pablo> 1.6 crop camera), it doesn't matter, PTStitcher will always
 Pablo> assume C, even if S is specified.

 Pablo> Since PTOptimizer seems to ignore S and C, and uses the input
 Pablo> image width for hfov calculation, the estimated parameters
 Pablo> will be wrong when used with PTStitcher/PTmender.

 Pablo> Actually one or two years ago, libpano12 didn't support
 Pablo> cropping circles that reach outside the source image, but
 Pablo> somebody added a patch to support it. Maybe things got broken
 Pablo> by it. Its strange that nobody ever noticed it.

Indeed, particularly given that several shareware products use the
library.

 Pablo> ciao


--
Daniel M. German
http://turingmachine.org/
http://silvernegative.com/
dmg (at) uvic (dot) ca
replace (at) with @ and (dot) with .

 


More information about the ptx mailing list