And about color temperature from a strictly RED ONE shooting POV...

[Emanuel A.]

Although, I'm not cinematographer yet. When I attended the film school, I was cinematography student during one month before my migration to the editing/directing fields. And I remember something about color temperature as one of the major points about the cinematic look. Crossing these boards, it's possible to read a lot of stuff about DOF, latitude, noise floor beyond those clichés already known as motion aka progressive image aka 24p or 25p, proper lighting, etc.

However, the color reproduction in terms of color temperature is not so common reported even if it is rarely mentioned, if ever mentioned.

I want to believe it's not merely a PP (post production) ability, so I'd like to read some ideas about colorimeter use and stuff related to one of the major issues IMHO to a major picture quality.

Any further input(s) ?

[GlennChan]

1- There's white balance. Sometimes people like to manipulate the white balance (to something ""incorrect"") to give a warmer or cooler image. In my opinion, this may not really turn out the way you want since your eye's "auto white balance" will mostly white balance it away.

Though on traditional video cameras, knee settings (or clipping) may muck around with the highlights so that they are a different color. So you get a tint in the image which has a different effect. But with Red, you can change all this stuff in Redcine since the image is recorded without image processing.

1b- Sometimes people warm up the image since it supposedly looks more like film. I'm not sure if that is true, but sometimes warming up the image does make it look better... though that's subjective.

2- There's mixed color temperatures in lighting. Sometimes this looks bad, like when you mix daylight from the sun or sky with tungsten lighting. The tungsten will appear very warm, or the daylight will appear very cool/blue depending on what you white balance to.

There are various lighting techniques to change color temperatures to what you want. Gel the light source, change the bulb, bounce light off colored surfaces (like gold reflectors), etc.

Some meters can measure color temperature, but they may or may not have issues with metamerism (Google that). A good meter would have similar spectral sensitivities to Red. It might possibly be simpler to carry a dSLR for scouting and/or use a field HD monitor.

[Lauri Kettunen]

I want to believe it's not merely a PP (post production) ability

The white balance is needed because our brain interprets colors in a special way as if the brain seeks all the time for white and gray. You can notice this effect in the winter during the sun set of a cloudy day. When you are indoors and have the lights on, you suddenly realize that the view on the window is blue. In Finnish we have a dedicated name for this: "the blue moment". However, if you rush outdoors to check the blue moment, as soon as you get your nose out you realize it's just dull grey there. What happens is, when being inside our brain interprets object colors according to the light generated, say, by the electric bulbs, which is yellowish. When you get out, the brain switches the interpretation according to the sun light filtered by the clouds and to the snow field, which the brain expects to be gray. Since a camera does not have the slightest idea of the object you are shooting, this effect has to be compensated by technical means, and this is what white balance is about.

Now, in Red One when the raw data is stored, it means you convert the voltage data of each pixel of the sensor to a digital number. That is, the raw data is about pixel voltages which correspond to the pixel light intensities. No colors exist at this point, yet. This is somewhat comparable to having a film which is not yet processed, but with the big difference that the film processing is a fixed procedure. Otherwise the film labs would drive photographers crazy with unpredictable processes resulting in arbitrary tones of colors. And once the film is processed, it can't be unreverted to get back to the stage of "raw" unprocessed film.

In case of RED raw data you don't have such restrictions. You can "process" and "revert" as many times as you ever like without loosing anything as long as you have the original raw file available.

For this reason, when shooting you don't need to worry about the white balance provided you store the raw file. But, if you store any other file format, then there's no way to revert, and then adjusting the colors become a post processing issue as you suggest. In post processing you also have less latitude to set the colors as you like. Thus, better to operate with the raw file.

[mdo]

White balance is a meaningless concept to raw sensor data. Different temperatures of light have energy curves that peak at different colors, with higher temperatures peaking at shorter wavelengths of light (more toward blue) and lower temperatures peaking at longer wavelengths (more toward red). White balance is simply a matter of factoring in how the current temperature of light is biasing the data that is coming off the sensor. It is done by taking a frame that contains white or neutral gray and calculating out how the raw data needs to be adjusted to make all color components equal. That is your white balance. There is no value in doing this in camera if you're capturing raw, as the white balance is not stored in that data. Just make sure that you get some shots of a white or neutral gray reference so you can easily balance later. If, on the other hand, you want to come out RGB, you would want to balance the whites, but recommendations on this site are to avoid coming out RGB if possible.

[jbeale]

Not sure if this is what Emanuel meant, but there is also the spectral color response of the eye vs. film vs. digital sensor. They are generally pretty close, but not exact. You sometimes see this (for example) when you take a digital picture of a flower that is a delicate purple shade. The resulting photograph may look less purple and more blue than the flower looked to your eye, even when the other colors in the scene do match. There are, or have been, portrait films made with four emulsion layers rather than just three to better capture skin tones. Color measurement and reproduction is a complicated field, I certainly don't understand all of it.

[Patrick Tresch]

I understood the theory... please clarify the "processing" of RAW material?

There is an issue about the color response of an object when lit with different color source.

I assume, the CMO should have different response to color when shifting from the "orinal value".

Common CCD have 3200k preset made for TV studios (thungsten lit) and then an optical filter is added to reach 5500k.

Today, the more used switch is an electronic one. You don't get the same result when putting 3200k preset and color correct electronically to 5500k using a white chart. You will lose sensitivity (color&light).

I don't understand how RAW data would "process", when there is no or very little response about a color how can you get a response in post by revealing it?

Thanks.

Patrick

[Graeme Nattress]

Patrick, if you use an optical filter, you loose sensitivity as optical filters can only reduce light, never increase it. With electronic white balance you don't loose sensitivity as electronic filters can gain and diminish.

We use a method for white balance similar to how Digital SLR's work, and how their raw conversion software works.

Graeme

[Patrick Tresch]

Yes of course, if you put an opitcal filter you cut ligth. I ment when changing electonically form 3200k to 5500k.

Pat

[GlennChan]

Some CCD/video cameras have optical filters to white balance the light before it hits the sensor. If you don't do this, then one of the channels will clip before the others. This can reduce your overall dynamic range. By using optical filtration, the channels clip at about the same level... so you get maximum dynamic range. In studio environments and in daylight, the reduced sensitivity is not a big deal since there is so much light.

In Red, Redcine should offer a highlight recovery option. So suppose you shoot without optical filtration. The channels do not clip at the same time. What the highlight recovery does is guess the missing channel information, based on the channels that haven't clipped yet. If the results suck, then you can turn highlight recovery off or dial it back.
This approach does not work in traditional video cameras / they don't implement it. (It'd likely be way too difficult to operate, especially when some of them have black and white viewfinders.)

Therefore if you shoot Red, using optical filtration doesn't necessarily make sense.


Not sure if this is what Emanuel meant, but there is also the spectral color response of the eye vs. film vs. digital sensor. They are generally pretty close, but not exact. You sometimes see this (for example) when you take a digital picture of a flower that is a delicate purple shade. The resulting photograph may look less purple and more blue than the flower looked to your eye, even when the other colors in the scene do match. There are, or have been, portrait films made with four emulsion layers rather than just three to better capture skin tones. Color measurement and reproduction is a complicated field, I certainly don't understand all of it.
That's a form of metamerism.

(If you assume that the 1931 CIE measurements and assumptions behind them are correct) then the ideal method of designing a camera is to have four color filters, with each filter corresponding to the CIE color matching functions (CMFs). The fourth filter handles the negative lobe of one of the CMFs. A color matrix will get you the rest of the way there.

Sony's RGB+E sensor design does something like this I believe... it's in their F828 still camera.

However, there are some problems with the 1931 CIE measurements. So that's where the color science gets murky.

Some sort of related information here:
http://www.filmlight.ltd.uk/documents/FL-TL-TN-0101-StdColourSpaces.pdf

Nonetheless, all current film+video cameras have color accuracy compromises but it's usually not a practical issue. We aren't good at spotting minor color inaccuracies.

[Graeme Nattress]

The electronic change is not done to your data. Your data is saved from the camera as raw data. The change is done in post. You can change the colour temperature to any number you like. You're not set to fixed presets. You can have any kelvin temperature from 1700 through to 100000, and there's a tint parameter to account for light temperatures that don't lie on the plankian locus.

Graeme

[Lauri Kettunen]

I don't understand how RAW data would "process", when there is no or very little response about a color how can you get a response in post by revealing it?

Patrick, do you mean, when the RAW file is converted say to a mov or avi file, how can one set the white balance if there is nothing in the image which can be considered neutral? If so, you mean, while shooting one needs somehow to set the color temperature, for example, with a white card? I guess you can do that and store the color temperature as RED metadata, but Graeme is the one to answer such a question.

[Graeme Nattress]

Yes, you white balance in camera and that gets stored in the metadata. Whenever I open up a RED Raw file, it comes in with the balance defined in camera. That gives you a starting point to do an adjustment, or you can find something white and "pick" or type in a known number for that white source.

Graeme

[Joel Kaye]

We use a method for white balance similar to how Digital SLR's work, and how their raw conversion software works.

Graeme

DSLR's usually have ISO settings. Does RED have something similar?

I'm really confused as to where and how light sensitivity could be adjusted. Is that all really just software manipulation of the raw data the sensor takes in? In which case boosting iso (or gain?) would be easier than decreasing light sensitivity (that would be very handy for bright sunlight situations).

[GlennChan]

Joel, your question may or may not warrant its own thread since it's not quite the original topic.

Anyways...
The Red camera can implement two different types of gain.
A- Analog gain on the signal before it hits the analog-digital convertor. (I think this feature will be on Red anyways; the Red team were posting test images up in the ISO threads.)
B- Digital gain by taking the digital values (1s and 0s) and multiplying them by some number. This would be the same idea as using levels in Photoshop.

As far as B goes, Red has an option to record digital gain information as metadata. The viewfinder will show the image with digital gain on it. But it is reversible since it's not baked into the data.


In which case boosting iso (or gain?) would be easier than decreasing light sensitivity (that would be very handy for bright sunlight situations).
In bright situations, you might want to lower the ISO/gain. If the camera does not lower it enough for you, you can add ND filters.

And of course you can adjust the iris, shutter, etc. etc.

[Patrick Tresch]

I asked a friend from AFP about Digital SLR!

Just find out what was possible to do with RAW. Amaizing!

Thanks Graeme for the your input!

Pat

[Michael Morlan]

"Planckian locus?" Thanks Graeme. Now I have something else I need to look up! :nerd:

http://en.wikipedia.org/wiki/Planckian_locus

[Graeme Nattress]

Can you see the lines crossing the locus in the diagram? These are iso-temp lines. Any colour not on the locus has a "correlated colour temperature" which you get by moving along an iso-temp line until the locus is reached. The distance the colour is away from the locus along the iso-temp line, is what we call "tint". By a combination of kelvin and tint controls, you can achieve any colour temperature for a white balance, but have it make sense in terms that we can understand.

Graeme

[Michael Hastings]

Relating to the idea of raw doesn't need white balance/color filtration, we underwater users have been thinking that would we would shoot with no filters. (Normally we white balance at depth and often use a U/W corrected red/orange filter to extend white balance range.)

Would there be a benefit in using the underwater filter given the extreme blue? Would this give us better dynamic range overall?

In short, what is your suggestion?





Patrick, if you use an optical filter, you loose sensitivity as optical filters can only reduce light, never increase it. With electronic white balance you don't loose sensitivity as electronic filters can gain and diminish.

We use a method for white balance similar to how Digital SLR's work, and how their raw conversion software works.

Graeme


Some CCD/video cameras have optical filters to white balance the light before it hits the sensor. If you don't do this, then one of the channels will clip before the others. This can reduce your overall dynamic range. By using optical filtration, the channels clip at about the same level... so you get maximum dynamic range. In studio environments and in daylight, the reduced sensitivity is not a big deal since there is so much light.

In Red, Redcine should offer a highlight recovery option. So suppose you shoot without optical filtration. The channels do not clip at the same time. What the highlight recovery does is guess the missing channel information, based on the channels that haven't clipped yet. If the results suck, then you can turn highlight recovery off or dial it back.Therefore if you shoot Red, using optical filtration doesn't necessarily make sense.


That's a form of metamerism.

[Graeme Nattress]

My suggestion would be to try it and see! I don't know enough about underwater light conditions to give a theoretical prediction.

Graeme

[Álex Montoya]

RAW is a whole new world. I've been experimenting with it recently with my SONY DSC-R1 and the extended latitude and overall control that it allows is astounding.

The simile with the photographic negative (RAW) and print (8-bit JPG, for example) is extremely accurate and enlightening. Adobe has its own RAW format that could, in the future, become a standard: they call it "digital negative" (DNG extension).

[Michael Hastings]

My suggestion would be to try it and see! I don't know enough about underwater light conditions to give a theoretical prediction.

Graeme

I was afraid you'd say that... that was the plan. As far as conditions, it's very blue, figure as if you were lighting a scene and the only light you had was a flood with a strong blue gel on it - would you want to offset that with a filter or just fix it redcine?

[Graeme Nattress]

Sensors are usually darker in blue than other colours, so too much blue might even help what's going on.

Graeme

[Daniel Reichenbach]

The problem is not, to correct one light (daylight, thungsten, neon, whatever) RAW will handle it, and I for myself are using since 20 years in 95% of the cases presets 3200 or 5600 (or 35mm/16mm D/T-Film), so I know, where my colours are. The problem is, when you have different lights from different sources, just to press a white balance knob won't help this much. You have to learn how to see different colours, much better than to learn, how to press colour balance buttons. :sarcasm:

[Lauri Kettunen]

In short, what is your suggestion?

Neither I can give a reliable suggestion, for have no experience of shooting underwater. But, there's somewhat the same problem in shooting the aurora borealis, so can give an example demonstrating what Graeme is saying and giving an idea how much the raw image can be adjusted.

Here's an image of the Northern lights taken with Canon 1D MarkII. The raw file is opened with different settings, color temperature 2000K and 50000K,
tint -150, 0, and 150.

Color temp = 2000K, tint=0:
http://www.reduser.net/forum/uploaded/3500_1188499895.jpg
www.luontovideo.net/CT2000-tint0.jpg

Color temp = 2000K, tint=-150:
http://www.reduser.net/forum/uploaded/3500_1188500371.jpg
www.luontovideo.net/CT2000-tint-150.jpg

Color temp = 2000K, tint=+150:
www.luontovideo.net/CT2000-tint+150.jpg

Color temp = 50000K, tint=0:
http://www.reduser.net/forum/uploaded/3500_1188500535.jpg
www.luontovideo.net/CT50000-tint0.jpg

Color temp = 50000K, tint=-150:
http://www.reduser.net/forum/uploaded/3500_1188500590.jpg
www.luontovideo.net/CT50000-tint-150.jpg

Color temp = 50000K, tint=+150:
www.luontovideo.net/CT50000+tint150.jpg

[Don King]

Would we get a better outcome using an external colourimeter or is it the same to only use the resources available in-camera shooting RAW?

Very interesting thread, actually.