[ptx] new reference for stitching algorithms

[Pablo d'Angelo]

Hello Aseem,

Very nice to hear for you! I haven't read the paper in full detail yet, 
but on a first glance it looks very good, brings together many
things that have previously usually considered on its own.

I think that the best way to solve the alpha problem (at least from our 
panoramic stiching standpoint) is the command line version. :) That 
could be used instantly (well almost, except for command line option 
changes) as a replacement for enblend :)

For heavier, later postprocessing tasks with your GUI it alpha channel 
support would be nice indeed. But probably you've used the wxWindows 
image loading functions that might throw away the alpha channel, which 
will need to be replaced by something else (and maybe throughout the code).

Do you have plans for further development of the software, or do you 
want to move on to other topics?

PS. I believe the "smartblend"
http://www.minorlogic.com/projects/smartblend/index.htm
software uses a very similar approach as your stitcher.

ciao
   Pablo

Aseem Agarwala wrote:

> The software currently can essentially operate on the same input as 
> enblend: images that are already aligned into one space.  However, there 
> is one caveat.  The software currently treats pure black as 
> "transparent."  If an image has an alpha channel, the software will just 
> ignore it, and treat the whole thing as opaque except where the pixels 
> are black.
> 
> The above is probably pretty annoying.  In the short term, I can offer 
> two solutions.  One, if you just want to see if the software has 
> anything to offer you, email me your most difficult and nasty stack of 
> 4-channel, aligned images, and I will stitch it for you.  Two, I have a 
> command-line only version of the stitcher that can handle 4-channel 
> images, so I could also GPL and release that.
> 
> Thanks for the interest,
> Aseem Agarwala
> http://agarwala.org
> 
> 
> 
> 
> 
> 
> 
>