Camera Profiles in V-RAY

[ARTECONI-CGI]

Hi All,

Happy new year!

I have been playing around with the LUTs recently and I was wondering if is there a way to correctly re-create the look of the best cameras available today.
I refer to Canon 5D Mark IV, Nikon 850 / 810, Sony A7rII/III etc etc..

I understand that if I shoot raw, with no profiles from the camera, the image is still not linear. Please, correct me if I am wrong.
Also, I would like to understand if the dynamic range of a camera affects somehow the image itself.
Which factor determines a Canon or a Nikon look? Do these cameras compress the highlights and crush the blacks by default?
Do they default to a sRGB profile?

Is there any way to reverse engineer these camera profiles and plug them back into the VFB?

Thanks!
Giacomo.

[joconnell]

- If you shoot raw, there's no way to view the raw as a developed photo - your raw viewer or image processing program is applying some settings to it.
- Dynamic range just means the camera can shoot a larger amount of exposure data on the high and low end without clipping
- most cameras will indeed put a contrast s curve on to the image in their develop settings (hence corona / fstorm doing the same) and the finer parts of the colour settings are down to the manufacturers tastes.
- Yep

You can use a still image of a macbeth chart shot by your camera which has known values. Bring that image into davinci resolve, nuke or 3d lut creator and I think all three have tools to take whatever image is inputted and colour correct them so that the macbeth values are the known true linear values. This makes a lut / colour correction that flattens your image. The three programs should have an option or a tool to do the reverse of the colour correction, or to apply the camera look to a linear image and you should be able to save this out as a .lut or .cube for vray.

[ARTECONI-CGI]

Hi Jo,

Thanks for the answers, it all makes sense.

What am I not completely sure is: How can a color profile be reverse engineered?

If I shot at a macbeth chart with a 5D Mark IV I will have 13 or more EV of exposure.
While if I shot at the same chart with an ixus 10 years old it will probably get less then 10 EV.
Still, the black will look black and the white will look white on both the images. More or less, or at least that's what I am guessing.

Therefore I assume that the 5D Mark IV will compress the highlight to fit in the same colour space. Unless the raw file format will take the EVs into account while creating the LUT...
Can you elaborate on this, it is a bit confusing for me.

[joconnell]

With raw files, you're going to get a little bit more data over white and a little bit under black. Any camera can capture the normal black to white range (which is from rgb 10 up to about rgb 230 so) and the ev / stops / exposure range of the camera tells you how much extra room above and below bright and dark the camera can handle before the data starts to get unusable. When you expose a raw file, it'll at some point have to clip the data to fit between our normal 0 black and 255 white but if you keep your raw file or convert to a floating point image (tiff / hdr / exr) you don't have to clip off the data in case you want to do more grading on the image after. For a monitor or printer though, white and black are your max points so any data outside of it can't be used. Likewise with your camera profiles they normally operate on the data between black and white - as you say though there might be some instructions in the raw file that rather than just going right to a burnt out highlight for values over white, it'll roll the values out a bit more gently - this is normally called "toe" for the low end and "shoulder" for the high end in traditional film exposure.

The macbeth chart doesn't deal with pure black and white values thankfully (no material in the real world exists with 0 black and 255 white) so any lut / calibration from this chart won't accurately deal with the top and bottom end - there's a bit of inaccuracy here normally so you can likely make some adjustments to taste here.

What you're asking about turns out to be really complicated to get totally accurate results by the way - I've been chasing it through some top end vfx companies and it normally gets to a "good enough" level and then the amount of effort required to get closer again might not be worth it. I haven't properly used it yet but rawdigger can take into account multiple exposures of the macbeth chart and even different lighting conditions. Grab a colour checker passport as it'll make some luts for your cameras for use in lightroom / photoshop and you can use it for all the bits above.

[ARTECONI-CGI]

Thanks again for your help, I'll try and shoot the chart in the weekend!

[kosso_olli]

This would be very handy, and something that I looked for for years. However, I have no idea how this can be achieved.

[ijzerman86]

I use the ones made by Kim Amland as mentioned in this thread.
Works best set to overlay on 68% intensity.
https://corona-renderer.com/forum/in...topic=13398.15
Also mentioned here:
http://www.peterguthrie.net/blog/

[kosso_olli]

Well, thats a good starting point. What I would like to have is different camera response curves from actual cameras.
I experimented a bit with this, and got an easy way to display the true linear image a camera sensor sees. But I have no idea of how to make a LUT of the response curves from this and the JPEG the camera calculates.

[joconnell]

Nuke can create a lut by comparing two images and outputting the color transform to a file, I think you can do the same in the free version of resolve too.

https://nofilmschool.com/2016/06/nee...-lut-can-do-it

[kosso_olli]

Thanks,
this looks good. So ideally I use dcraw to develop a "linear raw", a normal raw with camera raw and then compare the two like shown above?

[joconnell]

Yep that seems to be how dubcat on the corona forums has done the same!

[kosso_olli]

Okay, let me test around with this. Thanks for the help!

[Ihno]

John, I'd love to watch a proper curse from you about this whole (camera linearising / what do real cameras do to the image) topic.
I'm grabbing all those bits on the forum every time it pops up but I always have the feeling that I'm missing something. And it takes ages to learn it that way.
Do you know a good curse about all of this? I'd buy it right away.

[joconnell]

I don't yet but I'll be running through it with a new camera soon, when I get a lens to shoot panoramas on. I'll try and document the lot step by step and get something up.

[Ihno]

That would be awesome!