Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Dragon Non-Linear Color Response?

Wavelengths

Wavelengths

More and more shows/stages are moving to LED and other new school luminants. Their spectral distribution plots often show odd gaps and spikes at random wavelengths (probably at multiples or roots of the LED phase rate). All de-bayering algorithms synthesize colors based on the transmissive properties of the dyes in the color filter array. Given that...

Thesis: If the specific spectral reflectivity biases of a given makeup formulation happen to reinforce non-linearities in the output profile of the source fixture the bayer pattern reconstruction process may be "fooled" - especially if they match (or are out of phase with) the wavelengths of the dyes. I think this explains the anecdotal evidence, and, perhaps, even why its such a pronounced issue with the RED cameras. In designing the proprietary color science Graeme must have taken into account the character of the sensor and done some compensation (digital imaging 101). What if this created the equivalent of a "resonant frequency", essentially a band of synthesized wavelengths with non-linear reproduction issues.

Yes, in a roundabout way I'm blaming the sensor for requiring too much "help" in the design of the algorithm. The good news is that the Dragon optimized algorithm should be based on far more linear input values which should reduce the odds that any particular scenario would create strong color casts. Hopefully we are at the cusp of a new era at RED based on the purpose built Dragon sensor and the more linear color science it allows.

Cheers - #19
 
Remember remember remember.

Somewhere around the corner lurks DRAGONcolor and DRAGONgamma. Hopefully soon. Much of what all of "this" is won't be when that arrives unless you choose to shoot with RC3/RG3 still.

I feel the new OLPF and Calibration are also important keys here in terms of general image quality.

Yes, there are a good number of Dragons out there now. The 7 carbon Dragons here in South Korea are all working producing beautiful pictures, but what will be known as the "normal" Dragon image quality is yet to come.

Soon though. Very soon.
 
Blair S. Paulsen said:
More and more shows/stages are moving to LED and other new school luminants. Their spectral distribution plots often show odd gaps and spikes at random wavelengths (probably at multiples or roots of the LED phase rate). All de-bayering algorithms synthesize colors based on the transmissive properties of the dyes in the color filter array. Given that...

Thesis: If the specific spectral reflectivity biases of a given makeup formulation happen to reinforce non-linearities in the output profile of the source fixture the bayer pattern reconstruction process may be "fooled" - especially if they match (or are out of phase with) the wavelengths of the dyes. I think this explains the anecdotal evidence, and, perhaps, even why its such a pronounced issue with the RED cameras. In designing the proprietary color science Graeme must have taken into account the character of the sensor and done some compensation (digital imaging 101). What if this created the equivalent of a "resonant frequency", essentially a band of synthesized wavelengths with non-linear reproduction issues.

Yes, in a roundabout way I'm blaming the sensor for requiring too much "help" in the design of the algorithm. The good news is that the Dragon optimized algorithm should be based on far more linear input values which should reduce the odds that any particular scenario would create strong color casts. Hopefully we are at the cusp of a new era at RED based on the purpose built Dragon sensor and the more linear color science it allows.

Cheers - #19
Click to expand...

I don't use LEDs to light people. At least not on any of the jobs where I've seen the problem over the past 4 years.
 
Matt Ryan said:
Very true, but I've also never heard of anyone shoot HFR in available light only at night. Kind of shooting yourself in the foot with whatever camera you choose.

The Dragon ISO 2000 claim really threw me off too. Doesn't even seem to compare to 800 ISO on MX if they are exposed identically. I actually found the old calibration/olpf to do better in low light than the new calibration/olpf which shifted towards the highlights taking away from the low end. Wish there was a way for a temporal process similar to ADD to work in camera to reduce the noise and without further post processing. Maybe some on chip noise reduction.
Click to expand...

What's the difference between shooting a camera ISO800 at 24fps and one, that is supposed to be as noise free at ISO2000, at 48fps? Perhaps the reason you never heard of anyone shooting at 48fps in available urban night time light is because it was never before possible at 4K clean.

My thinking was that, since Dragon is supposed to be 1 1/2 stops better than MX in noise, it would be able to be pushed up one stop in shutter speed. That does not appear to be the case. From my little, non-film school experience: MX = Dragon in noise. Would like to hear otherwise or what I am doing wrong to get such noisy results.

Tonight I will see what 24fps results in. Probably not radically cleaner than 48fps. I'll also open up the Lomo the last stop to 2.8 which will compromise the sharpness no doubt.

If Canon offers a PL mount big sensor camera that shoots 48fps+ at NAB I will get one since I don't travel with a lighting crew for my work. Was hoping that Dragon would prevent me from carrying 2 cameras.

I get that current technology allows you a choice of 3 out of 4:
1. High ISO / Low Light Sensitivity
2. High Pixel Resolution
3. High Frame Rate
4. High Bit Depth

BTW: Respect to your body of work. An honor to see your comments and learn from you.
 
Gavin Greenwalt said:
Last night's Modern Family had some really really bad yellow makeup and it's Alexa.
Click to expand...

Sheesh. I saw that, too. For anyone who's interested, the episode was called "Las Vegas" and the makeup looked unusually bad. It's worth a watch and may be enlightening for those folks who think Alexa's shit don't stink. ;-)

To be fair, RED still has some work to do. The episode of Justified called "The Toll" has some pretty atrocious makeup happening on a few characters as well, especially the marshall that Rayland relieves at the end of the episode.
 
Zeb B said:
What's the difference between shooting a camera ISO800 at 24fps and one, that is supposed to be as noise free at ISO2000, at 48fps? Perhaps the reason you never heard of anyone shooting at 48fps in available urban night time light is because it was never before possible at 4K clean.

My thinking was that, since Dragon is supposed to be 1 1/2 stops better than MX in noise, it would be able to be pushed up one stop in shutter speed. That does not appear to be the case. From my little, non-film school experience: MX = Dragon in noise. Would like to hear otherwise or what I am doing wrong to get such noisy results.

Tonight I will see what 24fps results in. Probably not radically cleaner than 48fps. I'll also open up the Lomo the last stop to 2.8 which will compromise the sharpness no doubt.

If Canon offers a PL mount big sensor camera that shoots 48fps+ at NAB I will get one since I don't travel with a lighting crew for my work. Was hoping that Dragon would prevent me from carrying 2 cameras.

I get that current technology allows you a choice of 3 out of 4:
1. High ISO / Low Light Sensitivity
2. High Pixel Resolution
3. High Frame Rate
4. High Bit Depth

BTW: Respect to your body of work. An honor to see your comments and learn from you.
Click to expand...

Hey Zeb, my apologies. I wasn't criticizing you for your choice but rather pointing out that HFR shoots usually employ lots of light. If I was shooting available light footage I'd either go with a C500 or an F5 (most likely f5 for the internal recording option). Those should perform better in low light than Dragon. The F5 can shoot up to 60fps in 4K at the moment. I too am disappointed by the low light capability of Dragon (Although it is slightly better than MX) but then again it's pretty good considering the dynamic range and resolution of the sensor. I'm hoping Dragoncolor/DragonGamma improve on this.

Appreciate the compliment.
 
Gavin Greenwalt said:
Yes. I'm used to compression wonkiness I know what it looks like.
Click to expand...

OK, OK...

i wasn't so sure myself that is why I asked your opinion, but now I am still head scratchin'.

As I eluded earlier I make no effort to color calibrate the family TV because for the most part things look reasonably OK you know... Consistent, or at least how I expect them to look. Then some ugly green skin tones began appearing in House of Cards. It was likely coincidental, but when I looked closely at some problem images I noted macroblocks of cyan and magenta in the faces and that is what raised my questions about distro compression, decompression or perhaps upscaling issues. I never know what resolution is coming at me these days.

On the HOC thread I had read that some post work was happening in non-conventional places like airplanes and such. This makes me wonder if it might be a grading signal chain QC kind of thing at the heart of their seemingly inconsistent color.

It was also suggested that the production was very cognizant of the fact that a large portion of their audience would be viewing on mobile devices. The show does look more natural on the iPad and that has me wondering if they weren't targeting specific platforms to the detriment of others, but more importantly it tells me that the green color I was seeing was likely not a result of the photography.

With all of these questions I find it difficult to blame the camera for the skin tones I see on Netflix. There are too many other hands in the soup. I don't doubt that there is an issue with makeups out there, but to assume that we are seeing it in the final product of all of these programs casts doubt on the professional skills of a whole lot of people. Many DP's, Directors, Producers and Colorists would have to be screwing the pooch for that to happen. It doesn't seem likely to me.

I expect if I throw the HOC BluRay into my system that I will not see the same ugly green faces that Netflix streaming was showing me. I may just do that and see.
 
Last edited:
Just had a long conversation with a professional make up artist with experience in TV shows, feature films, TVCs, music videos, etc.


Summary:


1) Cameras have become very sensitive and their job has become more difficult. What was satisfactory before isn't any more. That even rhymes.

2) She cannot point out any particular camera for skin tone inconsistencies. Based on her set experience and seen masters (post part excluded).

3) Although make-up done in front of typical bulb-lit make-up mirror setups (and similar routes) looks cool for BTS videos, in reality it's less productive and reliable for these purposes, and it's suggested that light color/type used matches the one in front of the camera.

4) Make-up brand, if professional one, is less important. Application process and person applying it are the most important. (:eek:) - Sounds familiar ?
Her colleague worked on the same project with same make-up brand and got replaced due to poor skin look results. She didn't get replaced.

5) Airbrush is highly recommended. Much more even distribution, subtle nuances and more natural looking outcome.

6) She tends to mix her own colors, and tends to avoid using simplified color tones, especially ones sold for mass market, designed to be used for personal cosmetics and going out or typical photo shoots.

7) Color offsets she noticed were related to:
- certain skin enhancing types of lens filters which can bring unexpected results (she did not remember which)
- cheap or mixed lights

She did not make correlations to cameras.

8) Pointing out one or few colors/lines is very general and less helpful because a lot depends on how make-up is combined and applied.




Eric Haase said:
HAVE EXHIBITED THE PROBLEM: - -These are brands that typically lean towards yellow or golden.
Visiora (PL 006)
"HD makeup"
Bobby Brown
Makeup forever hd foundation



TESTED AND DO NOT EXHIBIT PROBLEM
Gerda Spillman (CH-8134)
La Mer Tan
Armani
Loreal Tru Match Cool
Click to expand...

First group

- Visiora & HD Makeup - no experience, although term "HD" in make-up is thrown around as marketing BS not guaranteeing anything
- Doesn't use Bobby Brown due to personal choice but has not had or heard any issues
- Make Up For Ever brand is widely used in this industry and is one of which she uses with no issues

Second group

- Gerda Spillman & La Mer Tan - no experience
- Armani & Loreal - have no place in professional make up. Loreal wants to enter this segment, results yet to be seen, don't believe the obvious and non-obvious marketing hype.


Her standpoint on cosmetics list approach, used in general as a reference, was that in practice it is less useful, for reasons listed above.
The exact words used were more harsh but I explained to her the premise behind the idea. Not to be taken personally, please. :)


Suggested brands:


Make Up For Ever
Make-Up Designory (yes, spelled DesignOry)
Mac Cosmetics


Not suggesting any certainties. This is one professional's experience and feedback and I suggest more research from fellow colleagues and comparing feedback.
If someone has particular questions regarding this segment I can ask her.
 
My interpretation of main causes of these types of issues in general:

- make up choice/artist/working conditions
- light amount/quality
- post work


Potential reasons why subtle nuances issues may not be perceived on the set:

- color environment and its effect on color perception
- eye sight is adaptive to priorities and focus, they differ on set and in post
- color precision in post once fully debayered


Potential reasons of debatable picture issues and related drama around Red cameras:

- large and widespread owner base, affecting types of projects, pushing the cameras far beyond optimal conditions than high end segment
- pricing segment & certain business choices related market perception, affecting production mentality > compromise range > final results, based on production budgets
- fast industry evolution, market expansion and market hype related unrealistic expectations, reactions to unmet expectations blown out of proportion through online noise


The best ways to screw up skin tone consistency:

a) Use make up from the nearest store, preferably powder, applied by anyone, with underexposed material, lit by source with spikes across the spectrum
b) Allow less experienced to fiddle with secondaries
c) Combine a+b
 
Thank you Hrvoje for the research! Anyone else care to place a data point on the chart of wisdom??
 
Hrvoje Simic said:
4) Brand of make up, if professional one, is less important. Application process and person applying it are the most important. (:eek:)
Her colleague worked on the same project with same make up brand and got replaced due to poor skin look results. She didn't get replaced.
Click to expand...

If you could, ask her about concealers and powders. It is my understanding that concealers in general are more opaque, more reflective, and that may be what the cameras are seeing.

I am particularly curious about powders, as some of them contain talc, and talc is highly reflective. If it is sticking to shiny/sticky areas and not so much to other areas, then inconsistent densities of reflective talc might be responsible in part for what we are seeing.
 
Points all well taken Hrvoje.

All I can say is we got green skin on a £2 sterling budget 5 week shoot with a very experienced make-up crew.

If you have seen this it is really obvious. It's on the LCD, not visible with the naked eye.

I would compare it to IR pollution in that respect (and I have suspected it may be related). The camera is seeing something the naked eye cannot.
 
Can't really add anything, but just wanted to throw in a reminder that the yellow you're seeing on your RGB monitors/screens, is all in your mind...

0:15 - 2:15

https://www.youtube.com/watch?v=R3unPcJDbCc

At least under sunlight our eyes are actually processing the real wavelength of light that our brains interpret as yellow. But with capture and presentation systems and signal chains that pretty much have to just infer the presence of certain wavelengths of light/colour, it's no wonder things go astray sometimes.
 
You must log in or register to reply here.
Back
Top