13 replies to this topic

[PeKi7i]

Is it posible to use diference in ungraded material and graded(primary CC) and to create LUT for primary CC for applying it on ungraded material?

[Tilt]

Not with the LUT cube fuses. But maybe the ColorCurve's match function does the trick?

How was the material graded? If it was done in something like lustre or another profesional color system... On a recent project I was able to match primary grading or printer lights exactly using a multiplication in linear color space. So just remove the gamma, add a ColorGain tool and use the advanced color picker macro from vfxpedia to sample the color of a certain area. Then adjust the gain sliders until the color matches the original plate.

[SirEdric]

Well...it should be possible...;-)
If you can make the CC's match curve (or any combination of color tools) to replicate the primary grading,
then you can take an instance of that CC (or color pipeline),
grade a LUTCubeCreator with it and utilize a LUTCubeAnalyzer to write out a 3DLUT for you
which then can be applied to raw material via the FileLUT tool e.g.

Cheers.

Eric.

[Tilt]

Yeah, Eric's right. You can get a LUT that way.
But the Fuses won't accept two images and calculate the difference, they need the LUTCubeCreator's test pattern.

[bfloch]

PeKi7i, on 14 November 2011 - 06:14 AM, said:

Is it posible to use diference in ungraded material and graded(primary CC) and to create LUT for primary CC for applying it on ungraded material?

Actually it is pretty obvious why not: a normal image only has a very tiny portion of the complete color space in it. So you could only match the difference for that particular subset of the colors that are used in the footage.

I actually wrote a (unaccelerated / time consuming) fuse for matching the difference but it's not worth it. You would need to extract the biggest possible color space from all the temporal information and still would need to guess / interpolate a big part of the LUT. Also any secondary grading obviously can lead to problems.

Bottom line: for safe results use the 1:1 LUT Cube. Please be aware, that the LUT Cube Creator is sampled on the B-Channel. Bigger size means more accurate blue channel representation. that you will get a more accurate representation of the color space if you increase the size of the cube.

Cheers,

Edited by bfloch, 14 November 2011 - 12:13 PM.

[PeKi7i]

Tanx guys for fast responses.So i need to manualy match the curves,i tought it would be possible to use color diference between graded/ungraded material for lut extraction(to be aplied on same ungraded material)

[ChadCapeland]

Eh? The cube is isotropic. You increase the resolution for R, G, and B equally.

[SirEdric]

PeKi7i, on 14 November 2011 - 10:11 AM, said:

Tanx guys for fast responses.So i need to manualy match the curves,i tought it would be possible to use color diference between graded/ungraded material for lut extraction(to be aplied on same ungraded material)

Noooo...not manually...;-)
Connect your raw material to the BG input of the CC.
Connect the graded material to the FG input
In the histogram tab click "match".
If the result is okay you can use a copy of that CC to grade a color cube as mentioned before.
(Along with the color-subset restrictions pointed out by Blazej)

Cheers.

Eric.

[PeKi7i]

Tanx Eric,I totally forgot CC histogram match(i never use it before to be honest) this is a fine tip
peace

[SirEdric]

You're welcome...;-)

Have Phun.

Eric.

[bfloch]

ChadCapeland, on 14 November 2011 - 10:13 AM, said:

Eh? The cube is isotropic. You increase the resolution for R, G, and B equally.

Yes you are right. Changed that. Thanks for pointing it out!

[ChadCapeland]

As to the color subset issue... CCv has the same issue. You have to figure out how to extrapolate that one way or another, but that extrapolation MIGHT be correct, at least enough to where it would matter, since it would only apply to colors that didn't exist in the original footage.

What's the correct grade for the mail box you just comped into this desert landscape? If the mail box is green, and your desert is just yellow sand and blue sky, who's to notice if the color comes out slightly wrong?

[bfloch]

It really depends on the source footage. Lets say you have a close up, extract the color collertion and then use that to drive a shot where new colors are introduced ... thats why I would love to have this extracting on time, which should provide more and more accurate results the more footage you analyze. If the grading was more or less similar.
For similar shots the technique single frame method is pretty ok once you use bigger thresholds for matching the colors.

Any sugestions for the extrapolation? I was thinking about turning the known points to csplines and and giving control over the whites and blacks.

[ChadCapeland]

We've got a new tool coming to the blog soon that does temporal color sampling and outputs a LUT Image capable of being used with the LCP. It won't do grading, but from a workflow standpoint, it does what we would like some of the color tools to do in Fusion, including proper masking (none of this rectangle business). BTW, we also updated fCMXto support LUT Images.