4
\$\begingroup\$

For many displays, we see the color space or gamut shown on chromaticity charts in comparison with some standard color spaces like Rec. 709, sRGB, Adobe RGB, and Prophoto RGB which gives an indication of the display device's color reproduction capability. Similar charts are are also done for many popular printers.

But not for DSLRs, like say the color gamut of a Nikon D300s DSLR plotted on a chromaticity chart? Why is this so?

\$\endgroup\$
1

3 Answers 3

6
\$\begingroup\$

The chromatic gamut is the property of output device, not a recording device.

The colorimetric input device is characterised with it's compliance to Maxwell-Ives criterion - i.e. how metamerically close it is to the CIE standard observer - the equivalent of photopic human vision. This compliance can be roughly characterised with one value, the example of it is DxOMark's metamerism score and metamerism is the most important property of colour recording device which should be characterised.

Theoretically, a recording device CAN have limited input chromatic gamut if, for example, it uses colour matching functions with negative responses present (I am just making it up because there is no practical possibility of doing this, a device doing this would be a broken device): CIE1931 CMF

If it happens so that electronics of that imaginary device are not capable of producing negative image data then wavelengths from 400 to 600 would be out of gamut for such device while some mixtures of wavelengths will be inside of gamut.

However, there is no obligation to use those functions with negative parts because there are completely positive colour matching functions which give accurate metamerism as well because when you have full chromatic data at all image points you can use arithmetic operations on all channels to deduce true colour: CIE XYZ CMF

No cameras I've seen have actual negative responses.

The most important thing - metamerism - gets broken in all consumer cameras regardless. You can't deduce true colours from the image data which a usual camera records. Here's an example of trying to deduce true colour from Nikon D70 data taken from http://theory.uchicago.edu/:Nikon D70 CIE best fit

This graph shows how well colours can be reproduced. Knowing that a CIE XYZ is a space of imaginary super-saturated colours you can see that colour reproduction accuracy is a trainwreck. And to top it off D70 image data gets clipped when transformed to XYZ space - which is in a sense the gamut limitation because XYZ is the widest colour space used after RAW processing.

So, there is no such thing as input gamut in the sense you are asking about.

\$\endgroup\$
10
  • \$\begingroup\$ > The chromatic gamut is the property of output device, not a recording device This is incorrect, the gamut is not specifically tied to an output device it can be the colours present in a scene: eilv.cie.co.at/term/211 \$\endgroup\$
    – Kel Solaar
    Aug 10, 2016 at 8:24
  • \$\begingroup\$ > it uses colour matching functions with negative responses present This is again incorrect, the matching functions you have linked are likely Stiles & Burch 1955 2° RGB CMFs r¯(λ),g¯(λ),b¯(λ) and were derived from colour matching experiments on human observers, the negative values are the result of the observers not being able to match the test stimulus with the mixture, thus one of the mixture component is added to the test stimulus and is considered as negative: github.com/colour-science/colour-notebooks/blob/master/… \$\endgroup\$
    – Kel Solaar
    Aug 10, 2016 at 8:32
  • \$\begingroup\$ It is important to acknowledge that those colour matching functions are not the cones responses of the human visual system nor sensor responses of any kind of capturing device. I think you are confusing too many things here. \$\endgroup\$
    – Kel Solaar
    Aug 10, 2016 at 8:33
  • \$\begingroup\$ @KelSolaar your reference doesn't say what you think it says. The gamut of an image is not the gamut of the recording device. A camera's gamut represents the limit of its capability, not the limit of reproducing some simple scene. Since the full color "palette" represents the neurological response to multiple-wavelength inputs, a digital image merely need match the cones' relative spectral responses. \$\endgroup\$ Aug 10, 2016 at 12:00
  • 2
    \$\begingroup\$ @kel-solaar: in practice cameras rarely have input gamut - i.e. the subset of human-visible colours which they can differ between.All cameras, with some rare exceptions, differentiate between every wavelength between 400 and 700. You are correct in that functions r¯(λ),g¯(λ),b¯(λ) were derived from experiments - that does not invalidate my statement, they are still valid CMFs which reproduce human metamerism. I never said that those reponses are human-like, I only said that they are equivalent because they reproduce human metamerism. That file requires some additional software. \$\endgroup\$ Aug 10, 2016 at 15:17
2
\$\begingroup\$

It is true that DSLR makers don't provide chromaticity diagrams for their cameras, they usually encode their images using sRGB or Adobe 1998 RGB colourspaces like @MirekE stated it.

Any colour lying outside those colourspaces gamut is clipped on casual RAW conversion, even if the camera captured it. It is also incorrect to think that RAW files don't have a gamut because as per CIE definition the colour gamut is:

[...] volume, area, or solid in a colour space, consisting of all those colours that are either:

(a) present in a specific scene, artwork, photograph, photomechanical, or other reproduction;
(b) capable of being created using a particular output device and/or medium

Because camera makers don't provide chromaticity diagrams for their DSLR doesn't mean they don't do it for their movie cameras, on those charts, all the gamuts represented except for the ACES family ones and Pointer are from movie camera vendors (ARRI, Canon, Sony, Panasonic):

ACES Encodings - Diagrams

If you are interested on how to characterise digital camera and how their colour gamut are computed, Martınez-Verdu et al. in Concerning the calculation of the color gamut in a digital camera has a lot of relevant information on the topic.

\$\endgroup\$
13
  • 1
    \$\begingroup\$ Camera sensors simply accumulate charge. Every pixel is just a range of intensity, so a RAW file really doesn't have color per-se, nor does it really have a gamut. The color of the data in a RAW file is entirely a matter of interpretation by processing software. Even if you tag the RAW file with sRGB or AdobeRGB, one is not required to use those color spaces for processing. It's just metadata. The pixel data is limited by noise and the full well capacity of the pixels, and there may be some slight non-linearity in the extremes of the range...but, RAW is RAW. \$\endgroup\$
    – jrista
    Aug 10, 2016 at 17:44
  • 1
    \$\begingroup\$ The color filter array over the sensor will determine the exact response of each pixel to specific ranges of the spectrum. The curves for any given sensor are going to depend on the specific nature of the color filters and the specific nature of the silicon's response to light. So the sensor itself (not the RAW image file) will have a unique response, however most fall within a given range as unfiltered silicon itself has a fairly consistent response to light, ranging from ~315nm up through ~1100nm, with the response low in UV, growing through the visible spectrum, and peaking in IR. \$\endgroup\$
    – jrista
    Aug 10, 2016 at 17:48
  • 1
    \$\begingroup\$ So the sensor itself may have a gamut (although I don't think that is the correct term...the sensor has a natural response to light)...but I don't think it is correct to say that a RAW file has gamut. \$\endgroup\$
    – jrista
    Aug 10, 2016 at 17:49
  • 1
    \$\begingroup\$ I honestly think you are reading the definition to suit your view of "gamut", the definition first subject is a scene, the scene being what you are trying to reproduce. I'll let you dig the references from here: github.com/colour-science/colour/tree/develop/colour/models/rgb/…. \$\endgroup\$
    – Kel Solaar
    Aug 11, 2016 at 10:51
  • 3
    \$\begingroup\$ Practically speaking, any unique color that fails to produce a signal greater than the noise floor for a given sensor can be said to be out of the sensor's gamut if the same amount of that color is present in full spectrum white light at an exposure value sufficient to reach full well capacity of R,G, and B sensels. \$\endgroup\$
    – Michael C
    Aug 15, 2016 at 3:09
1
\$\begingroup\$

Cameras can shoot either in RAW, sRGB or AdobeRGB. For sRGB and AdobeRGB the gamut is given. RAW files do not have gamut, you need to convert the image to some color representation like RGB to get gamut and that is a process that happens outside of the camera and the gamut will depend on the editor, profiles and the taste of the person who operates the editor.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.