Aviator technicolour process...

Hugh · **Aviator technicolour process... -** January 29th, 2005, 07:11 AM

Salatar showed me this page which details some of the colour process that The Aviator went through....

He challenged me to replicate this and post a tut on how to do it, part of which I have done, and the other part I am in the process of doing now....

The video is quite vague on how to generate the mattes for each channel.... Here is how to do this:

The maths for generating the base for the mattes is:

Red: Red - ((Green + Blue) / 2)
Green: Green - ((Red + Blue) / 2)
Blue: Blue - ((Red + Green) / 2)

These then need to be inverted and 'expanded'

In Shake, I found the best value for the Expand node was a low value of 0.8 and a high value of 1. For those of you not using Shake, this means that what used to be at 0.8 is now at 0, and what used to be at 1 is now at 1. The values in between are stretched accordingly....

To build the final image, you need to multiply the original red channel by the average of the green and blue mattes, and so on and so forth.

Doing it like this, I got the following result"

Original image:

My converted image:

Their converted image:

Mine is slightly different, but a lot of this (the blue of the sky, for example) can be tweaked in the Expand nodes....

I've also attached the Shake script that I used to do this to this post.

LFGabel · January 29th, 2005, 05:50 PM

Nice! Thanks, Hugh!

jkast82 · January 29th, 2005, 07:01 PM

Hugh - thats excellent - cheers man

jimbo · January 30th, 2005, 06:09 AM

tanx
excellent tip

caretz · 11:03 AM

man U did a great job!!!
I have a question ....
do U think is possible to get close or more manageable result using a color correction node ?
I understand the nature of the experiment is just a curiosity...
beside I apreciate the effort that matches greatly the result

SalaTar · February 4th, 2005, 12:03 PM

It was fun

dfx · 09:33 AM

SalaTar interesting flow.

You can streamline it by using hugh's formula in Digital Fusion custom node.

r1-(g1+b1)/2
g1-(r1+b1)/2
b1-(r1+G1)/2

then invert using color space "to color" negative

then use brightnes and contrast node and set Low threshold to eg 0.5364238 and high to eg 0.9072847

then another custom tool with 2 inputs. input 1 being the orig clip.
r1*(g2*b2)
g1*(r2*b2)
b1*(r2*g2)

hope this is clear

dfx · February 5th, 2005, 09:45 AM

Anyone figured out the Two strip color process in a compositing app?

I'm Having problem converting RGB to CMYK in Shake or Fusion? (not CMY in compositing apps which is just an invert of rgb)

however, I Can do the 2 strip process in Photoshop converting RGB to print CMYK and manipulate the channels as shown in the Aviator demo.

One standard equation that is used to convert RGB to CMYK is:
Black = minimum(1-Red,1-Green,1-Blue)
Cyan = (1-Red-Black)/(1-Black)
Magenta = (1-Green-Black)/(1-Black)
Yellow = (1-Blue-Black)/(1-Black)

no idea how to do that in Fusion

hugh any ideas?

http://www.neuro.sfc.keio.ac.jp/~aly...space-faq.html

arlen · February 5th, 2005, 09:46 AM

does anybody have any idea how to do it in CB, it seem quite different, i got really funny result with isolate RGB and 1/2 channel mattes together

SalaTar · 04:09 PM

I like the luma key sliders for tweaking

If you apply Hughes math to I real image ...what happens?

dfx · February 5th, 2005, 10:56 PM

Ok i quit

All nite trying to mimick 2 Strip Technicolor and i have to sleep.. Wife is pissed cuz i stayed in office for 2 days now

shhh

Anyone know how to mimick Photoshop print CMYK in Fusion?

In that pic, how would you get from "New Red Layer" to Cyan and "New Green" to Magenta in Fusion?

Easy to do in photoshop.

Quoted from the site:
"This multilayered matte strategy was constructed in After Effects and produced a generic "Technicolor Filter" that could be applied to any image including a multicolored chip chart. This chip chart produced the basis of a 3D LUT by comparing each color before the "Technicolor filter" and after."

Whats a 3D Lut? Somthing like a color match ?

Hugh · February 6th, 2005, 06:42 AM

I'm thinking that you get to the Cyan by doing the original - R

I'm going to have another play with this one...

Patrick Sheffield · **Maybe math error? -** February 6th, 2005, 11:19 AM

The math being used is:

Red: Red - ((Green + Blue) / 2)
Green: Green - ((Red + Blue) / 2)
Blue: Blue - ((Red + Green) / 2)

I think I see a problem in the script - the adds were taking place before the divides and the mattes were peaking out/clipping.

It should be:

Red: Red - Green/2 + Blue / 2
Green: Green - Red/2 + Blue/ 2
Blue: Blue - Red/2 + Green / 2

Re-ordering the nodes to do the Mult before the IAdd should fix this.

Patrick

Hugh · February 6th, 2005, 01:50 PM

Thanks for posting that in here, Patrick - I'm on the Shake list, but don't seem to be able to post to it at the moment...

If I remember right, I converted the image to float before doing anything else, specifically to stop it clipping like you're describing.... I'll have to check if I managed to only think this part (I don't think Shake has a thought control plugin yet... I'm working on one, but it's a bit flakey....)

Patrick Sheffield · **You're right -** February 6th, 2005, 02:15 PM

You are correct, you did. The reason I started down this path was that I was looking at the intermediate steps and it looked in the viewer like the mattes were clipping, so I changed the ordering and the mattes looked better in the intermediate stages. I haven't had time to do a full test on it. I was unsure of the math following the subtract where you clamp, invert, expand, add, multiply before multiplying them with the original mattes. Why can't you simply combine the mattes after the initial subtraction? I'm just trying to wrap my head around the theory...

Thanks for doing this, it's really quite fun.

Patrick