Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1084)
Subject: Re: [AE] Inverse a LUT?
From: Brendan Bolles <brendan@fnordware.com>
In-Reply-To: <list-4615181@media-motion.tv>
Date: Thu, 2 Feb 2012 19:32:03 -0800
Content-Transfer-Encoding: quoted-printable
Message-Id: <BBF10C3B-AFB6-46AA-9001-581B5DAF0FF9@fnordware.com>
References: <list-4615181@media-motion.tv>
To: "After Effects Mail List" <AE-List@media-motion.tv>

On Feb 2, 2012, at 6:48 PM, Chris Zwar wrote:

> Not sure if other software such as Nuke can do this, you might have to =
render out 32bit dpx or exr and have them converted back to Arri LogC in =
a different package.

I don't know of any software that can invert a 3D LUT.  This is a =
problem you're going to have with any film look.  You just have to try =
your best to integrate your elements with standard color correction =
tools.

Here's another way to look at the impossibility of inverting a 3D LUT: =
we were working on the movie Hellboy, which has a big blue glowing cloud =
(the "hell hole") at one point.  The director kept on saying he wanted =
the thing to be more saturated, but no matter what we did we couldn't =
get the blue as saturated as he wanted.  If you applied one of these =
film previews, no matter what color you put in your log file, you just =
couldn't get the saturated blue that you could pick through the regular =
color picker.

Why?  Because that color blue was out of gamut for the film being =
previewed.  A film print was simply not capable of producing a blue as =
saturated as the director wanted.  Our inability to produce a saturated =
blue through the film look was telling us that.  So if the director were =
to pick a saturated blue in Photoshop and tell us to put it on screen, =
it just wasn't possible.

If 3D LUTs could be inverted, it would be possible to pick any output =
color and get back the required source color.  But it's not possible.


Now, you could make a 3D LUT using a simple gamma operation.  And then =
another 3D LUT made with the inverse gamma would in fact be its inverse. =
 But I know of no software that can figure out if a 3D LUT is invertible =
and then actually do it.  Sounds like a good Ph.D thesis to me.

This might be worth a try though: convert a 3D LUT into a 1D RGB LUT, =
and then try to invert that (this can be done with LUT Buddy, I think).  =
Might get you part way there.


Brendan