Color measurements on RED?

[I Bloom]

Along the same vain of the spot meter and the RGB histogram, I'm wondering what sort of color measurement tools we'll have access to in camera. Chroma meters are scarce and expensive and it would be great to have something that could help us easily measure color temperature, green or magenta shift in lights, correct for florescents and etc.

White balance is part of the equation, but in some situations we need more info to match mixed lighting.

So could there someday be a vectorscope and or some kind of spot chroma meter?

IBloom

[jbeale]

Part of the expense of a color meter is the calibration to a known standard reference. Are video cameras normally calibrated in that way? Would that add time & expense to production?

[Craig Schober]

Part of the expense of a color meter is the calibration to a known standard reference. Are video cameras normally calibrated in that way? Would that add time & expense to production?


Yes, currently balanced around 5000k. I'd not shoot with a filter though - I'd use the highlight tool in Redcine if you were to have a colour of light that caused a shift in the highlights (and as the colour gets corrected from un-clipped channels, you get some extra guestimated highlight detail too).

If you shoot with a filter, the kelvin white balance numbers won't make sense any more....

Graeme

if you know the red cam is 5000k, then you should be able to assign a number to any other chroma values in camera. if they haven't already built something that will do this in camera, i'm sure it will take some time and additional expense but it doesn't seem impossible for this feature to be added as a firmware update down the line.

[jbeale]

I'd be all for it as a firmware feature that gives a quick way of comparing light color temps. For many productions I assume relative consistency is more important than an absolute reading (or more likely, a "correlated color temperature" if it's not a true blackbody).

If I was doing it though, I'd want to make clear whether such readings as "around 5000k" means 5000k or 4700k ... 5250k, etc. If just you give people a number they often assume it's accurate, and then they complain if their expensive calibrated Brand X meter doesn't agree.

As I recall my DVX100 displayed the focus distance in arbitrary "units" rather than feet or meters. I assume it was because they didn't want to hear complaints about the accuracy (and also it would be wrong if you added a wide adaptor).

[I Bloom]

if you know the red cam is 5000k, then you should be able to assign a number to any other chroma values in camera. if they haven't already built something that will do this in camera, i'm sure it will take some time and additional expense but it doesn't seem impossible for this feature to be added as a firmware update down the line.

Well I think what Graeme is saying makes some sense only if you need the Kelvin numbers in the camera. I think if you are shooting a gray card through a color correction filter it should work the same as film. It just depends on if you think you are getting better quality with a filter or by shifting in post.

I'm thinking more about correcting lights for example by adding small amount of plus or minus green. And trying to detect things like HMI's or Flos that aren't quite right.

In that sense we really just need some numbers that give us an average reading within a small box at the center of the sensor. That way we can place a graycard there and see how far it is from gray. Then maybe in camera it can convert from RGB to:

Tristimulus values: XYZ
Chromaticity: Ev xy, Ev u'v'
Correlated color temperature: Ev, Tcp, uv
Color difference: (XYZ), (Ev xy), (Ev u'v'), Ev u'v'

Really anything will do. Just to get some numbers, instead of trying to guess about color based on the calibration of monitors.

I'm imagining that at the beggining of a shoot we might just set up the camera and test out some of our lights, see how things are burning, maybe cook up some gel concoctions. Then on set if we see something weird, we can flip on the color meter toss a graycard in the frame and see how its going.

IBloom

[Nils Ruinet]

I agree, a kind of vectorscope or chroma meter would really be useful.

[GlennChan]

I think partly why Graeme is telling you that you may want to avoid using CC filters (over the lens) is because Redcine will do chromatic adaptation when you white balance.

When you look at a scene, your eye will mostly white balance to it and everything looks right. But there can be some subtle differences in color depending on illumination. The spacing between just noticeable differences (JNDs) changes... where a JND is how far two tones have to be apart before you can just notice that they are different. One example would be consumer CRTs, which tend to have a very blue white point... reds look a little de-saturated when the white point is so blue, so TVs will intentionally oversaturate reds so that it looks more correct (red push).

So when you shoot on Red and you white balance in Redcine, it will try to mathematically model things like that.

Some related math/color science info:
http://brucelindbloom.com/ChromAdaptEval.html

*Granted, this is just the theory behind it... I don't have much real experience in playing with it.

2- The other reason between filtering and not filtering Graeme already mentioned (and it may be more important to you). They are two different methods for maximizing dynamic range.

3- I think an easy method to accomplish what you're looking for is to simply look on a monitor. If you need to exaggerate the mixed color temperatures, then crank the chroma up on the monitor.

3b- A slightly different approach would be if Redcine (or the camera) displayed what the white point was in objective terms... e.g. chromaticity co-ordinates, and/or present white point on the Kelvin scale with an indication of how much +- green/magenta.