Re: [PanoTools-devel] Support for circular fish eyes

[Daniel M. G. <dm...@uv...>]

 Erik> On Monday, November 20, 2006 at 12:26, Daniel M. German wrote:
 Erik> If it works like in PTStitcher leave it like that. People and other 
 Erik> software are used to it even if it's not logic...
 >> 
 >> One side effect of the code I added for circular fisheyes is that if
 >> the images contain a mask it might be overwritten (if it is ouside the
 >> region that PTmender considers to be outside the region-of-interest).

 Erik> As far as I tested PTSTitcher works like this: The crop circle 
 Erik> diameter is adjusted equal to the longer side of the crop area and 
 Erik> centered in the crop area. If no crop is given, the image dimensions 
 Erik> are taken as crop. 

 Erik> If there is an alpha channel it is treated as mask and only pixels 
 Erik> inside this mask are processed. If an image is marked as circular 
 Erik> fisheye pixels outside the image circle are treated as not present, 
 Erik> regardless whether they are actually there, masked or unmasked.
 Erik> If PTStitcher does feathering it does this on pixels only that are 
 Erik> available after cropping and after masking. PTStitcher never includes 
 Erik> pixels that where excluded either by a mask or by a crop.

Yes, that is the way PTmender is operating. It will be nice if one of
you with Circ. fisheyes runs a project through it to make sure you get
expected results. I am particularly interested in those cases in which
the images are presented in landscape mode to panotools (in the
abscense of crop the horizontal width of the image is used to create
the mask).

I agree with you Erik and Pablo. There is no need to mess with the
masks. They can be replaced in the original image instead. 



--
Daniel M. German                  "The World Wide Web looks more
   The Economist ->                like the World Wide Mess."
http://turingmachine.org/
http://silvernegative.com/
dmg (at) uvic (dot) ca
replace (at) with @ and (dot) with .