search

Psyop / Cryptomatte

Cryptomatte Nuke plugin, Fusion plugin, sample images, and specification
BSD 3-Clause "New" or "Revised" License
637 stars 152 forks source link

[Nuke] 2.4 Gamma masks? #57

Closed dgirondi closed 6 years ago

dgirondi commented 6 years ago

I've been noticing a tiny little fringe around the edges of the generated masks from Nuke's Cryptomatte node in VRay renders that seems to be fixable by gamma correcting the masks to 2.4.

Is that expected behavior or am I doing something completely wrong here?

Renders are from Maya's Vray (Lin float), Nuke 10.5v6 and the latest Cryptomatte master release.

I've just checked with a test image posted here by another user and the same things seems to happen there (file attached).

cryptomatte_2 4-gamma

Cheers, Diogo

jonahfriedman commented 6 years ago

Hi @dgirondi,

I think all you're doing there is expanding the matte a little bit. You had a white pillow over blue here, try switching their colors and doing the color correction on the pillow, you may find you want to gamma down instead of up.

If you want to make absolutely sure VRay is doing the right thing, render an RGB matte of the pillow as well and compare the edge pixel values.

dgirondi commented 6 years ago

Exactly, using either 2.4 or 1/2.4 tends to fix the problem. I just got interest on the magic number.

I have a job here where we have the exactly same element with both cryptomatte and multimatte and I'll do further comparisons.

Thanks!

From: jonahfriedman notifications@github.com Sent: Tuesday, December 19, 2017 12:59:02 PM To: Psyop/Cryptomatte Cc: Diogo Girondi; Mention Subject: Re: [Psyop/Cryptomatte] [Nuke] 2.4 Gamma masks? (#57)

Hi @dgirondihttps://github.com/dgirondi,

I think all you're doing there is expanding the matte a little bit. You had a white pillow over blue here, try switching their colors and doing the color correction on the pillow, you may find you want to gamma down instead of up.

If you want to make absolutely sure VRay is doing the right thing, render an RGB matte of the pillow as well and compare the edge pixel values.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/Psyop/Cryptomatte/issues/57#issuecomment-352782782, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AAfs4k0IOalJTgDi1kNEV5K43Xx5kuEOks5tB882gaJpZM4RGor5.

jonahfriedman commented 6 years ago

Cool, thanks. I kind of doubt there's anything special about that number.

If there is an unexplained mismatch between the multimattes and cryptomatte output, that'd be at ticket for Chaos Group, but as far as I know there's no problem there.

Thanks