Dragon Non-Linear Color Response?

[Zeb B]

So I'm starting to grade the "Dragon first look" shots I took a couple weeks ago and this scene is making me track the model's face to subtract green. Is it me or maybe the sensor/color science loses sensitivity to red as the luminance increases?

I assure you the model's face is the same skin tone as the rest of her body. R3D's and below JPG's are of a nude woman with all the Garden of Eden lady bits showing so maybe NSFW

https://dl.dropboxusercontent.com/u/117794729/A004_C001_0223GG.0000469F.R3D

https://dl.dropboxusercontent.com/u/117794729/A004_C001_0223GG.0000607F.R3D

 

[Antony Newman]

Guess : Compensations for MX sensor on Dragon footage might have this effect.

AJ

 

[Eric Haase]

What kind and brand of makeup is she wearing? I have had significant problems with the MX sensor with certain types of makeup. I've posted over the years about these issues. The problem is most egregious under tungsten (or warmer) light. Yellow/orange base makeup is the culprit. That's kind of what this looks like to me. I was hoping this would go away with the Dragon sensor, and I'll wait to do my own tests before passing judgment. The workaround for me has been to tell the makeup artist to use more pink tones, even when counterintuitive to the person's skin tone. Or to use 5500K light or an 82C filter under tungsten. The latter are less frequently options when the problem arises by surprise.

 

[Zeb B]

"Hey Girls . . . put on this Hobbit make-up . . . it's for the Dragon sensor" Really?!?

 

[Gavin Greenwalt]

Yeah when I was looking at the wedges Phil posted earlier, it was pretty clear that even the RLF output wasn't truly linear out of Dragon yet.

 

[Hrvoje Simic]

I assure you the model's face is the same skin tone as the rest of her body.

What about the make up difference ?

I suggest same tests with different light source and without the make up to remove those factors out of the equation.

 

[Graeme Nattress]

If the sensor actually measured like that Gavin, we'd have really visible issues on any decent gradient. I'd strongly suspect that it's a measurement error as I've not been able to duplicate that kind of response and I've measured more Dragon sensors for linearity than I ever really wanted to do....

Zeb - I'll take a look and see if I can see anything.....

Graeme

 

[Phil Holland]

If the sensor actually measured like that Gavin, we'd have really visible issues on any decent gradient. I'd strongly suspect that it's a measurement error as I've not been able to duplicate that kind of response and I've measured more Dragon sensors for linearity than I ever really wanted to do....

Zeb - I'll take a look and see if I can see anything.....

Graeme

Correct there was a slight variance in that measurement/test shot on the prototype I had a while back which was mentioned in the old thread. I mentioned I'd redo this test in the future, which I'll do when I'm back in town.

 

[Zeb B]

If the sensor actually measured like that Gavin, we'd have really visible issues on any decent gradient. I'd strongly suspect that it's a measurement error as I've not been able to duplicate that kind of response and I've measured more Dragon sensors for linearity than I ever really wanted to do....

Zeb - I'll take a look and see if I can see anything.....

Graeme

Thank you Sir

 

[Timothy Eaton]

I have had the same or similar issues as Eric with yellow based makeup under tungsten light on Epic. Certain areas of the face begin to exhibit an unnatural yellow cast that's very difficult to remove. Under most circumstances it's just not an issue. In a few cases it was troublesome.

 

[Zeb B]

The swinging light bulb was a 60w antique style big filament tungsten. Almost looks fluorescent. Seems counter intuitive that putting all that orange light on a face would result in the opposite color reflecting back. Weird science man!

 

[Brian Boyer]

What kind and brand of makeup is she wearing? I have had significant problems with the MX sensor with certain types of makeup. I've posted over the years about these issues. The problem is most egregious under tungsten (or warmer) light. Yellow/orange base makeup is the culprit. That's kind of what this looks like to me. I was hoping this would go away with the Dragon sensor, and I'll wait to do my own tests before passing judgment. The workaround for me has been to tell the makeup artist to use more pink tones, even when counterintuitive to the person's skin tone. Or to use 5500K light or an 82C filter under tungsten. The latter are less frequently options when the problem arises by surprise.

I have had the same or similar issues as Eric with yellow based makeup under tungsten light on Epic. Certain areas of the face begin to exhibit an unnatural yellow cast that's very difficult to remove. Under most circumstances it's just not an issue. In a few cases it was troublesome.

The MX sensor has gotten a bad rap for the blotchy, yellow makeup issue, and probably rightly so. It does seem to be more susceptible to it than other sensors.

However, I do think a makeup's brand/color/formulation/application as well as the type of lighting when applied (is there a standard or does it vary? i.e. incandescent vs. CFL) and shot (tungsten vs. daylight) plays a part, whether it's captured digitally or on film.

I've noticed pretty severe makeup issues on NBC's The Blacklist, which is shot on the Sony F55.

I also just saw a pretty bad issue on TNT's Perception (Season 2, Episode 14, Obsession), which I believe is shot on the Alexa (according to Arri's website).

And, just for kicks, below is an image from Cloud Atlas, which was shot on film, showing that yellow blotchiness on the forehead.


Right-click for a larger image.

 

[Timur Civan]

What kind and brand of makeup is she wearing? I have had significant problems with the MX sensor with certain types of makeup. I've posted over the years about these issues. The problem is most egregious under tungsten (or warmer) light. Yellow/orange base makeup is the culprit. That's kind of what this looks like to me. I was hoping this would go away with the Dragon sensor, and I'll wait to do my own tests before passing judgment. The workaround for me has been to tell the makeup artist to use more pink tones, even when counterintuitive to the person's skin tone. Or to use 5500K light or an 82C filter under tungsten. The latter are less frequently options when the problem arises by surprise.

Yes! I just had this problem out of nowhere on a shoot this week. Model looked like cheetos were rubbed all over her face...

 

[Eric Haase]

I've been talking about this for a few years and a long time ago Graeme offered to help with tests but I've just not had the opportunity to do them with a proper makeup artist. This continues to be a big issue that pops up only occasionally but really wreaks havoc on the usually tight shooting schedule. The problem I have found is with yellow/orange base makeup. "HD" makeup, beard concealer, and a host of makeup brands and tones cause faces to look green and orange/green. You can frequently see the line where one tone ends and the other begins. The rest of the body looks magenta in comparison. I run into this issue when we have talent that comes from the broadcast world or sitcoms and have their own makeup artist. I have started a list of problematic brands.

I have shot side by side with a 5D, and the 5D does not render the same problem as the MX sensor. I have not encountered the problem with the Alexa.

 

[Hrvoje Simic]

I've seen make up issues on a couple of digital cameras, including few cases with Alexa.
In most cases it looked like the combination of lighting and make up light reflection/absorption properties.

And, just for kicks, below is an image from Cloud Atlas, which was shot on film, showing that yellow blotchiness on the forehead.


Right-click for a larger image.

Jim Broadbent has that type of skin tone variation, which is likely additionally affected in post.


http://www.aceshowbiz.com/events/Jim Broadbent/jim-broadbent-premiere-cloud-atlas-05.html

 

[Brian Boyer]

Jim Broadbent has that type of skin tone variation, which is likely additionally affected in post.


http://www.aceshowbiz.com/events/Jim Broadbent/jim-broadbent-premiere-cloud-atlas-05.html

I looked at this pic as well as several dozen more on IMDB. He does seem to have some natural yellow variations in his skin but it's hard to tell to what degree. As they're all shots of acting related or public appearances and, being an older gentleman, he's almost certainly "made up" in all of them to some extent.

It could be the case that whoever does his makeup never bothers to apply it higher than his forehead. After all, he is 6' 2".

In the pic below (from Another Year), you can see where the makeup stops at the top of his forehead, leaving the top of his head a slightly different color and less shiny.

I think whatever makeup issues a camera has is often exacerbated by not applying makeup close enough to the (actual) hairline or not blending well/far enough into the neck.

I notice a lot of women in general have this issue when wearing makeup - their faces will be considerably darker or lighter (depending on complexion) than their neck and chest. It's not a look I'm fond of.

http://ia.media-imdb.com/images/M/M...nXkFtZTcwNTUzNTE3Mw@@._V1__SX2298_SY1261_.jpg

 

[Zeb B]

K5600 Lights

K5600 Lights

NSFW

Same model, same make-up . . . different lights. Used a single K5600 HID on this scene. Doesn't look like she's wearing organic peanut butter face paint here . . . although you can still see the roots of the issue.

If the Dragon is a blank sheet new design, then how can it still have the random yellow/brown make-up with tungsten light color issue? Is this a to be fixed with the new Dragon Color Science deal? Impossible to see this issue during a small team session.

https://dl.dropboxusercontent.com/u/117794729/A005_C014_02239I.K5600HIDlight.R3D

Thanks for the help and advice

 

[Blair S. Paulsen]

Once we have mature Dragon color science, it would be great to create a database of spectral reflectivity metrics of widely used make up products. Zeb's example may have other issues, but knowing what to look out for with various types of make up could be a real asset. We've had some extensive threads on skin tones, but not that many posts about the spectral reflectivity characteristics of make up.

Photographers and make up artists have been collaborating for decades to achieve the intended "look". With film rapidly disappearing as a capture medium and the speed of change in digital capture so rapid, perhaps we need to make a greater effort to coordinate the selection of the make up chemistry, application techniques, etc with the profile of our capture systems.

Cheers - #19

 

[John Marchant]

My thoughts on this.

First up, RC3 / RG3 is not optimal for Dragon. Fine tuning to correct for this is doubling up on inaccuracies, as you are fighting inherent corrections that RC3/RG3 has for MX. Improvement on this from the ground up will be better than anything you can easily achieve with RCX tools.

Yellow blotchiness under tungsten is pretty easily produced by an under-lit over compressed blue channel. We're very sensitive to skin tone, and subtle variation in the poor cousin blue channel under tungsten can produce exactly this effect.

If you're into testing, check out the difference in skintone accuracy between an exposure rated ISO320 and ISO800 under tungsten, then maybe push to an extreme and rate the camera 3200... compare on top of that variation in compression ratio. Lower compression and lower base ISO has a substantial effect on this phenomenon.

IMHO Dragon predominantly improves on MX in this regard by being cleaner in the blue, and allowing for smaller compression artefacts (higher res) and improved compression in general.

 

[Zeb B]

What if all your have is 60 watts of light, a T2 lens with a shutter as open as possible? ISO Rating doesn't make a diff then . . . does it? I mean at the RAW level . . . it is what it is in a low light situation