LogC workflow question

[martind]

Hi all,

I've have a question related to the logC color managment. I've read the paper "Alexa LogC Curve - Usage in VFX" and I have a question about tone mapping and Rec 709 conversion.
Acording to the paper: after aplying the LogC curve there is a tone-mapping and the final color space conversion (from Alexa wide gamut).
There are two options for output color space in the Alexa web lut_generator:

Color space - none: 1D-LUT
Color space - Rec 709: 3D-LUT

So I have two LUTs generated by the web lut_generator:

AlexaV3_K1S1_LogC2Video_EE_nuke1d.cube
AlexaV3_K1S1_LogC2Video_Rec709_EE_nuke3d.cube

The 3D LUT combines tone-mapping + color space conversion in one 3D-LUT and the 1D-LUT includes only tone-mapping (correct?).
There is a color matrix described in the paper for conversion from tone-mapped ALEXA wide gamut RGB into ITU Rec. 709 RGB (1.485007 ...).

So if I understand the paper correclty, following workflows should generate same results?

A. linear source image -> Alexa logC curve -> 3D-LUT
B. linear source image -> Alexa logC curve -> 1D-LUT -> Rec 709 color matrix conversion (1.485007 ...)

But the results are different (tested in Nuke). It looks like there is a more complicated color conversion from Alexa wide gamut to Rec 709 than the simple color matrix mutliplication provided in the paper...

Thank you for the clarification,
Martin

[Harald Brendel]

Hello,

The conversion steps are
Log C -> Tone Map -> Display Linearization -> RGB Matrix conversion -> Display Encoding

You can not apply a color matrix to non-linear display data. The tone map LUT you get from the LUT converter yields an image for display on a standard video monitor. Five years ago there were some systems around that could apply a 1D LUT only. The tone map LUT has a gamma correction for 2.4 built in. You have to undo this gamma correction and apply after the matrix.

Hence, when your display is a standard video monitor (ITU Rec 1886)
Log C -> Tone Map -> x^2.4 -> RGB Matrix conversion -> x^(1/2.4)

This should get you pretty close to the 3DLUT. In the camera (and in the 3DLUT) we do not use a simple power function but something that approximates the power function with a limited slope around 0. It shouldn't matter when you do your processing in 16 bits ore more.

Best

[martind]

Thank you, it works.

We sometimes need an inverse LUT (texture and matte painting), and for the 3D-LUT it can't be computed.
I've tested the 1D-LUT in Nuke and it is close enough for our purposes...
Thanks again