Divide blending mode issue in VFB2

[piotrus3333]

Im trying x/(1-x) operation to reverse Reinhard tone mapping x/(x+1).
with the divide I'm getting more of an inverse than divide with a layer stack like that:

[Zdravko.Keremidchiev]

Could you please elaborate a bit more? Reinhard color mapping is a blend between Exponential and Linear multiply.
Also, Nuke produces the same result with that composition.

[piotrus3333]

sure.
so I’m talking about Exponential. and with the stack from the image I’m trying to go back from exponential to linear. but folder with ’divide’ does not seem to calculate 1-rgb before dividing.

[piotrus3333]

to clarify:
exponential color mapping saved from VRay
linear multiply saved from VRay
fusion comp with the math
output from fusion - linear from exponential:

[yavor.rubenov]

Hello,

To get x/(1-x) - you'll need to put them the other way around. First the folder that does (1-RGB) and then the RGB with divide mode:


Best regards,
Yavor

[piotrus3333]

Hello,

To get x/(1-x) - you'll need to put them the other way around. First the folder that does (1-RGB) and then the RGB with divide mode:

Best regards,
Yavor

Thank you, much appreciated.

[Zdravko.Keremidchiev]

I’m trying to go back from exponential to linear

This math won't work as the exponential mode clamps all values above 1 and you can't comp it back to linear.
https://docs.chaosgroup.com/display/...aultParameters

[piotrus3333]

well it seems to be working just fine. it’s not comped back to linear but „decompressed”. I checked it with Fusion and now with VFB. I still do not get why the layer order in VFB needs to be different (reversed?) though.

the exact math here is (1-RGB*0.99) - to avoid black pixels where exponential color mapping reaches 1.0 (self illumination in this case):