Hello--just getting into oiio dev and running into a strange issue with
deep data.
I'm seeing different sampled values between what oiio deep_value returns
and what the Nuke sampler returns. My simple test case involves a sphere
with RGB values between 0.0-1.0 over the surface, solid alpha (resulting in
1 sample/pixel), rendered to exr single scanline half float.
Piping the image through Nuke's deepRead->deepSample returns different
values than deepdata.deep_value(p,c,s) for the same pixel sample. For some
pixels, the value of a channel is a match. For other pixels, the value of a
channel is off by a small margin of ~.1. Oddly, the green channel seems
more affected/inaccurate than other channels, for some pixels even
returning a different sign, which makes me wonder if there's a color space
issue at play here.
I have not tested on the stable release (1.6) or any other branch. Any
ideas?
oiio version 1.7 (latest dev commit from 13 days ago)
Nuke version 9.0v5
exr ilmimf 2.2.0
ilmbase 2.2.0
Windows x64
Visual Studio 2013
Regards,
Mike
--
.......:::::: Michael Breymann
www.glyphfx.com ::::::.......
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
