> I have yet to see your definition of what constitutes a color space

"Color Space: A three-dimensional space or model into which the three attributes of a color can be represented, plotted, or recorded. These attributes are usually called hue, value, and chroma." "More commonly, color spaces are built on external references, such as the CIE system of measurement. "

Technically, a camera sensor is sensitive to some range of frequencies, which includes infra-red (and probably x-rays). The filters block certain frequencies and there are a set of grayscale numbers output which represent the data as seen through the red, green, and blue filters. Those data are linear (counting photons), which isn't useful until they are mapped into a well-defined color space, wherein the relationships among similar colors are handled consistently. It is then useful to convert from one color space to another, because every printer, monitor, etc. has a restricted range of what colors, saturation, and intensities/luminosity can be displayed. When you try to map a wide colorspace into a narrow one, then "out of gamut" colors are undefined. Convert your favorite photo to the smallest colorspace: one pencil and paper. Artists do it, but they make a lot of decisions about what gets left white (paper) and what gets a gray or black pencil. Even smaller: Adobe Photoshop "threshold" maps to black or white, no grays. The only decision is the cut-off point.

The camera color space is NOT well-defined without knowing the spectral response curves, etc. In these discussions. most people refer to well-defined color spaces: sRGB, Adobe RGB, LAB, ProPhoto .

Interesting thread, not!! Exhausting and tedious and much more than I ever need or want to know now or never, ever. You guys all need to go out and drink some strong drinks, pinch some women and then come back, trashed, for the fifteenth round. Don King would not be proud of this excersize in mind bending, latent color space grappling caused by testoserone leakage directly into all of your brains. Stuff just like this is what caused his hair to stand up straight. Right about now I think Rodney is looking in the mirror and wondering what the hell happened.

My raw files have a color space...They are inside of a powerful button on my mouse. I click it and sparks fly, smoke churns out of my fan port and I'm glad the boys at the workshop figured it out for me cause I don't want to know anything more about it. I've got enough to keep me busy, what with my teenagers out drinking and racing stolen busses down Ashland Avenue.

Keep up the good work and don't buy tube socks at K-mart.

Such discussions are useful because they remind us just how important definitions are. Many arguments can be resolved simply by getting an agreement on definitions. One you've got agreement on a satisfactory definition, it's then much easier to determine if the conditions of a particular scenario are consistent with the definition.

I could have interminable discussions with the Pope for example as to the definition of atheism, but I don't know much about color spaces .

I want to give my contribution.

Some definitions:

- gamut = gamma, or range, of colors
- color space = a reference for values that fix a color
- color = human sensation

Generally color space and gamut term are used as synonyms. So let me do the same..

Camera try to imitate the human vision.

1. Humans has 3 cones (red, green, blue) sensitive to visibile light.
1.Cameras has 3 filters (red,green,blue) sensitive to visibile light (frequencies out of visibile are filtered out)

2. the stimuli are conveyed to the brain that process them, discount, more or less, the illuminant color and yields the color sensation

2. the stimuli and the white balalance info are recorded for raw processing (in camera or out of camera). But spectral response of camera sensor is not.


Now, can we speak of “human gamut” ?
Yes the colors humans can see were standardized from CIE in the far 1931, The well known horseshoe-shaped area in the CIE -xy chromaticity diagram is the human gamut.

Can we speak of “camera gamut” ?
I think so. Any camera has a gamma of colors that it can see. I think this gamut is similar to human gamut, as a camera try to imitate humans,

Are the info recorded in a raw file enough to produce a colored image?
The answer is no. Sensor spectral response is not recorded. So you have to know it from a different source.


Now. Is a raw file grayscale or colored?

Grayscale definition: a single value builded fusing color info with some algorithm. There is no way to rebuild original color from a grayscale value.

In a raw file you find red green and blue values.
Yes the raw is not ready for rendering until it is demosaiced and converted to some known color space. We can say that a raw must be processed to became a color image… but

A jpg file must be processed to became a color image.
A tif or png file is not ready for rendering until you know what the values mean (image color space). It must be processed to became a color image

So, in digital, every image file must be processed.

To discrimate grayscale file from colored file we have not to think if a process is needed to extract color, but we have to think if is possible to extract colors.

A grayscale file has no color info, so we can’t build colors.
A raw file has color info, so we can build colors.

Jacopo

[quote=Angst,Nov 26 2007, 03:05 AM]
Schewe has it spot on!
1. Noise comes from high ISO (sensor) long before the color space gets applied. A wider gamut simply stretchs it, but doesn't add to nor subtract from it. To cure noise, shoot at ISO 100 or maybe 200. At 400 or beyond, you are going to have to deal with the noise.

The reason that I started this thread was that I was convinced that setting Prophoto in camera raw instead of Adobe rgb I could see more chroma noise and concluded that the wider colour space was the culprit Obviously I was wrong at least acording to Schew No offense meant Still I learned a lot more about colour spaces Some of it useful and some of it not!

Gamut isn't gamma if that's what your trying to say here. Gamut don't equal gamma either. Gamma is a simple mathematical formula.



Fix a color?

This is all about using the proper language to discuss a technical topic.

Monochrome:
Monochrome \Mon"o*chrome\, n. [Gr. ? of one color; ? single + ?
color: cf. F. monochrome.]
A painting or drawing in a single color; a picture made with
a single color.

Well there's color, ONE.

Now looks look at color spaces of which you say a Raw file has (even tough you admit its a Grayscale file)



Further, its useful to read the work of Poynton (http://www.poynton.com/notes/colour_and_gamma/ColorFAQ.html#RTFToC7)



Do you see anything referring to monochrome, defining color?

You agree that a Raw file is Grayscale. What does Apple in its developer's guide have to say (http://developer.apple.com/dev/techsupport/insidemac/ACI/ACI-48.html):



Note that this piece refers to the ColorSync manager and how it handles different color spaces yet they define gray as gray spaces NOT gray color spaces for what should now be obvious reasons. They are very precise and careful in defying these spaces and when they use the term color:




Further (http://en.wikipedia.org/wiki/Monochrome):


Lastly, let me remind you and others of an old Chinese proverb: The first step towards genius is calling things by their proper name.

You are trying to be scientific, I'm trying to be colloquial.
But you have to read well post. I said "gamma of colors", not "gamma".

My english is not so good, but I can try to explain "fix a color".

When you say that RGB values are (120,50,70) and you say that values are in sRGB color space, you are "fixing" what is the color you are referring to.

"setting" is more appropiate than "fixing" ?

Jacopo

By your definition, the raw file is not monochrome, because it has information for more than one color. The raw file has only luminance information. However, if we know that a certain luminance value is from a pixel overlain with a green filter, we know that it represents green luminance. The same applies to the other primaries. If we know the characteristics of the filters, then we can apply a 3 by 3 matrix transform to get the XYZ tristimulus values (Poynton discusses this). The XYZ space lacks a white value, but the raw file contains white balance information. Also we have a defined tone curve with a gamma of one. We have elements of a color space, depending on how a color space is defined. Unfortunately, you have yet to divulge your definition of a color space, so further debate is futile. Your tactic is to withhold your definition of a color space, and trickle out objections as needed as to why a given set of parameters is not a color space.

Jeff supplied a link on Wikipedia to absolute color spaces. This implies that all color spaces are not absolute. He seemed to imply that the camera space was not absolute, and hence not really a color space. This contradicts his own reference.

Reduced to absurdity, you can't get blood from a turnip, and you can't get color without color information, which may be explicit or implicit. Why don't you just give it up?

Your style of quotes is suboptimal. The first quote shows me as the source, but the others are anonymous and one can not judge their authority. Some are from Wikipedia and others are selected according to whether they support your view point or not.

I think you may mean gamut, not gamma no? So you mean gamut of colors, not gamma of colors.



Ah OK I see. What I'd say is this is the scale of the numbers. That's what a color space defines.

R0/G255/B0 can be used to define the most saturated green numerically using a common set of numbers found in Photoshop to define color. But these are different colors in sRGB versus Adobe RGB (1998) because the scale of the numbers fall in different locations within human vision.

So, if I say my car gets 23 to the gallon, I'm missing a scale (miles or kilometers). Or if I say, the price of gas in my neighborhood is $3, I'm not suppling the scale (gallons or litters).

What a color space does is define the scale of the numbers. It think that's what you were referring to as "fix". One could say fixing a scale to the numbers, how's that?

You understood perfectly now. Sorry for my english.

Jacopo

Nope, it doesn't contain color information, the entire point of this debate.



We do?



If. But again, you're still looking at this as what COULD be produced from the existing data, not what the existing data is.

A Raw file is no more a color image than a moth is a butterfly. Or an CMYK file is an RGB file just because we converted the color space. Lets stick to what the data is at a fixed point in time. As I said previously of course a Raw file will become a color file. But it isn't one yet.



I probably should. But you get so close to 'getting it' by admitting that a Raw is a Grayscale file but then jumping to what it will be after a process and then saying its a colored image, I keep thinking you're getting close.



That so many sources dismiss your premise for one. The first was from Wikipedia as well as the one other. I provided you a source from Poynton and Apple. But if you're going to disagree with all the sources provided that dismiss your points because I failed to provide you a link, I think you're bottoming out here in making any kind of valid point.

Did anything I posted not fully describe the difference between a Grayscale, monochrome file and a color file? You said Raw was Grayscale (after starting all this by saying my post stating that Raw is Grayscale was incorrect) yes?

Yes or no, is a Raw file Grayscale data?

Yes or no, did anything I posted from the various sites agree that a Grayscale file is a color file?

Can you find anything other than a vague mention in the DNG spec that says a Raw (Grayscale file) is a color file which I submit it doesn't say? Not what it may be, what it is NOW?

Wishful thinking on your part that Wikipedia, Poynton, etc dismiss my points. I beg to differ. Whether a raw file is gray scale or not is semantics. It does contain tri-stimulus color information or we wouldn't get color. The DNG spec does not deal with these nit picking semantics (which are irrelevant to the discussion), but is hardly vague concerning mapping the Camera Color Space to the CIE XYZ space. It does not define a color space, but implicitly recognizes that the camera does have a space. Why argue any further, since this is very clear to anyone who does not apply Clintonian logic. Now, Bill Clintion is very smart, but you can't pin him down and he is very good at parsing a phrase. However, in the end he was impeached.

You can do that, but you can't convince anyone else of your points if you continue to make this a religious not scientific discussion. As such, I will now say we are done here. If you want to believe that a Grayscale file is now a color file or that the earth was created in 6 days, despite carbon testing and sound science, or wish to dismiss the writings of Poynton and others, fine.


OK.


http://www.pmel.org/Color-Glossary.htm:
CIE Tristimulus Values: Amounts (in percentages)O of the three components necessary in a three-color additive mixture required for matching a color: in the CIE System, they are designated as X, Y and Z. The illuminant and standard observer color matching functions used must be designated; if they are not the assumption is made that the values are for the 1931 CIE 2° Standard Observer and Illuminant C.

Once you have a color file, you can assume the color space which I've said over and over again (or it can be specified). And again, one can describe the moth that will become a butterfly, but until that process, its still a moth.

What camera color space? Where is this defined within the Raw which isn't in color?
And again, this happens when you actually HAVE a color file that does contain tristimulus color information. But you refuse to accept this so be it.

Oh boy, you really are stretching here. Yup, I think its best we let this die. Go ahead, have the last word. You don't want to go down in LL forum history with that silly quote.

I was gonna let this one slide but then I remembered that the ONLY way to deal with the Epson 2000p was to make custom profiles for the Epson based upon the viewing light _BECAUSE_ of the metameric failure. In fact ProfileMaker was the first to put in the ability to make profiles based on the illuminant because inkjet inks at the time had metameric failure so bad. A profile made under tungsten sucked under daylight and vis versa. So, one single profile never worked well with that printer.

Which is also the reason that Thomas profiles a sensor at D65 and Standard Illuminate A and one of the reasons it can't be said that a camera has a fixed set of tristimulus values (spectral response) because the actual values depend on the spectral response, which will vary.