Gamut of RAW files

I think experimentation on this matter is out of my league.  

 

The disturbing thing about this is how easy it is to find opinions on color spaces but how hard it is to find the real data for hardware.  Maybe I'm looking in the wrong place.

Is this http://www.kodak.com/global/plugins/acroba...1CELongSpec.pdf

an example of what you're looking for? See page 12.

Not all image sensor makers publish these specs, however.

read entire topic!

brain gamut over loaded :shock:

must dive and reboot!

CB

It's close, Herb. I wish the color response curves would be converted into a plot in a color gamut much like films are. Here is a site wtih examples of that. You need to navigate to the "Beta RGB" section under the Info tab. General color space discussions are also found there.

If we knew this information we could be certain what color space would be sufficient.

Craig,

Doesn't the spec from Kodak give you enough info to do it yourself? There are raw converters that will give the uninterpolated output. You can take that and do the interpolation yourself. I think this is where the gamut mapping takes place.

I doesn't give me enough information to do that, although it may be enough for someone else.

Sifting through source code is time consuming and assumes that (1) the code is correct, and (2) you know the color model of the target. I'm not certain that (1) is true and (2) is definitely not for me. One thing is for certain, there is a color model already in place in the RAW file. It is not "created" at conversion, it is translated.

Herb, would you interpret the Kodak information as having a wide gamut, limited gamut or what and why would you say that? If the gamut of the device is wide, then I'd rather work with linear-mode, non-color managed TIFF's. I see no reason to munge data and throw a bunch of it away and the suggestion is that this is exactly what occurs when we choose a color space. Dynamic range is compromised when we apply the assocaited gamma as well. So this is my real motivation. I want to know if it is best to pursue a linear workflow to avoid discarding data unnecessarily.

"Input devices (scanners and digital cameras) don't have a color gamut because there is no sharp boundary between colors that they can "see" and colors that they can't - no matter what you put in from of them, they're going to see something. Instead, we say they have a color mixing function, the unique mixture of red, green, and blue values that they will produce for each color sample."

Compliments of Fraser, Murphy and Bunting, "Real World Color Management" Chapter 2

Digital cameras don't have a fixed gamut. They do have a fixed dynamic range.

They do go into more detail on color spaces, etc. but I kind of glazed over when staring at 3d charts.

I believe the gist of it is, that because the camera has a wider dynamic range than you output devices (monitor or printer) it can produce values outside of the output devices color space. When the value is outside of the color space, it then uses some method of translating into something withing the color space. The wider the color gamut of the output device, the less the need for this translation, which is a good thing.

Of course, I could be completely wrong.

I'd lke to know how they define color gamut then. A tristimulous device may see all colors but alias them into a subset of all colors. It's that subset I want to know.

I'm not sure how dynamic range fits into the concept of a color space. There are certainly integer encoding issues that effect the granularity.

Anyway, this is a good lead. Thanks, George. I've ordered my copy. I really want to understand what's going on here.

It would be great to have a tool that analyzes and image and maps it onto the the color space associated with the selected profile. Then we can see how much of a wide space like Adobe RGB or ProPhoto RGB is really used.

Craig,

OK, I give up. What does "tristimulous" mean?

This is an old thread, but I came across this article:

http://www.nikondigital.org/articles/rgb_d...amera_color.htm

Toward the bottom there's a reference to a camera's native color space which the author claims is typically called its "color gamut". Unfortunately, it doesn't answer the question here, but it does provide some details on the response of the D1x's bayer filter. Wonder how similar modern cameras are to this?

As much as I dislike referring to a Luminous Landscape article, this one pretty much addresses the matter. The quick conclusion is that digital sensors have very broad color response and ProPhoto RGB is needed to contain their output. AdobeRGB is not large enough.

I know this thread is old as the hills. But, I just found wetpixel.

Actually the software does quite a bit in determining the coloring of raw files. Based on the discussion it would appear that everyone has a good idea about the Bayer array in front of your camera's sensor. If not there is a good description of it here http://www.cambridgeincolour.com/tutorials...era-sensors.htm

Each individual pixel determine it's level of light however, the actual color of the pixel is not determined until the software evaluates the pixels around it. Only then can the actual color of that pixel be determined. So in the strictest of terms, the sensor has no color gamut. It is the entire system; bayer filter, sensor and software interpreting the sensor data that determine the gamut. On most, if not all, digital cameras the WB data is included with the raw file simply as reference. No WB correction is applied to the raw sensor data. When the raw file is viewed with software, the software has a profile for that camera and that raw file. Through interrupting this, the colors of the pixels are determined and the raw file can be manipulated for color.

I was recently going through some old photos from a dive trip to Tahiti. I had some raw images that were practically black. After some heavy lifting I was able to pull a decent looking photo of a leaf fish from the dead. The photo, not the fish. The fist was very alive.

That's a matter of semantics. I've found experts who refuse to use "gamut" in that context and others who are fine with it. There's no doubt, though, that digital sensors have a range of spectrum that they respond to and it's usually broad. Whether you want to refer to that range of response as gamut or something else is another matter. My personal feeling is that gamut, "the complete range or scope of something", is an appropriate word to use.

Software doesn't contribute to the range of color that the sensor responds to, it effects what you are capable of getting in the color-managed output file. Raw converters aren't necessarily accurate and visually distinct colors can map to the same sensor values, so there's some art and some science to the process. Clearly, software has an impact the "output-referred" gamut but that was never my question.