5
\$\begingroup\$

Recently on a wildlife photography course, the leader asserted that using aRGB in the camera settings meant less loss of data in the RAW file than if sRGB were used. As I did not understand this I checked on the web, finding flatly contradictory views. Eventually, I found this question which seemed to be THE definitive answer. Then investigating the EXIF contents of some of my recent shots, taken on a Nikon camera, I found that the choice of colour space was stored in the RAW file.

So my questions are:

Is the colour space stored in the RAW file?

If it is stored, is it true that using sRGB causes information to be lost, compared to using aRGB?

If it is stored, why does it not matter which you choose (as far as the RAW file is concerned)?

If it is not stored, why is the choice stored in the EXIF data?

I am aware that choice of colour space is relevant at the point at which the NEF file is converted, but my question is specific to the RAW file only.

\$\endgroup\$

2 Answers 2

6
\$\begingroup\$

Part of this question basically seems to be asking if the answers to the other question are actually right. For that part, don't worry — they are. (And no information is lost in this way — the leader of your course is wrong. *)

The other parts of the question basically ask if (and why) this information is stored at all, if it doesn't affect the raw data.

The answer is: like many camera settings other than the fundamental exposure factors (shutter speed, aperture, and sensor gain/iso), the color space selected is stored as advisory information, so that you can make choices in-camera and have them reflected as defaults in RAW processing software. This might not make a lot of sense with color space, but consider tone curves (or even black and white mode), white balance, and even crop. None of this is baked into or affects the basic RAW data itself, but all can be useful to have later.

Keep in mind that Adobe RGB is not strictly better than sRGB — that's a common myth. This is especially true in 8-bit images, like JPEGs. See What's the difference between Adobe RGB and sRGB and which should I set in my camera? for more — but, again, don't worry about it when using RAW. A greater concern is that your camera's rear LCD (or EVF, in mirrorless) is likely to be calibrated to show sRGB correctly, and using Adobe RBG may actually result in less correspondence between what you see there and what you're actually capturing.


* except possibly with the very minor theoretical effect you might see from possible biased metering, as explained in my answer to that other question.

\$\endgroup\$
3
\$\begingroup\$

In Brief:

ONE) The "leader" was incorrect, in fact he's got it backwards in a way. A RAW file is RAW image data and a colorspace is NOT RAW it's debayered. A chosen colospace has no effect on the RAW data, only the embedded jpeg preview image in the RAW file.

TWO) Please do not use aRGB to describe Adobe98 if that was what you meant. "A" implies an alpha-channel, not a colorspace, and Adobe is a brandname, it should be capitalized. The s of sRGB means "standard." And for other reasons, aRGB is bad form...

THREE) "Loss of data" has absolutely nothing to do with colorspaces, and everything to do with BIT DEPTH.

FOUR) There is a greater loss of image fidelity with larger colorspaces like AdobeRGB unless the bit depth is substantially increased. The only thing a larger colorspace does is put more empty space between code values. With an 8 bit image that means you can rapidly cause the deltaE errors to exceed 2, which means artifacts become visible (i.e. banding). The larger the colorspace, the more bits you need per pixel.

FIVE) sRGB is an ideal colorspace for 8 bit image presentation. And if you are not clipping, such as bright saturated colors, you will gain ZERO benefit from using a larger colorspace.

SIX) But none of that has anything to do with RAW, which is image data right off the sensor, before debayering. Image data before debayering does not relate to a "colorspace" as it is RAW data that has not been through a matrix, LUT, or other algorithm to transform the RAW samples into an RGB image. (Though you can and should create a profile of that sensor's RAW capture data, but that is not the same as a "colorspace".)

\$\endgroup\$
4
  • \$\begingroup\$ Well, technically RAW has a color space, it is defined by the color filters. It is even a wide gamut color space, similar to XYZ. But since it is invariant, and the raw processing software already knows about it, there is no point in encoding it in the metadata. \$\endgroup\$ Jun 2, 2019 at 12:26
  • \$\begingroup\$ Hi @JeroenvanDuyn — yes, that is what I meant regarding the profile. However, I consider it incorrect or misleading to call the RAW profile a "colorspace" because it is data that has not been debayered, and thus the image is not "constructed" and the colors are not "defined". By whcih I mean that before debayering, the raw values from the filters do not relate to definable colors, as the debaying takes image structure into account before any given pixel is assigned a color. Color spaces like XYZ define a specific color based on tristimulus values. RAW does not (continued) \$\endgroup\$
    – Myndex
    Jun 2, 2019 at 18:11
  • \$\begingroup\$ In RAW, a pixel has only one of three (or four) colors from the bayer filter. Surrounding pixels must be taken into account for the image to be constructed, and the image thus must be constructed to determine the color value of a given pixel. I suppose it is somewhat a semantic argument, but a colorspace can be converted to another space by a matrix or a LUT. RAW data requires a more complicated algorithm specific to the camera/sensor. Essentially I'm making the argument that a colorspace is generally independent of a specific device, and a profile is generally specific to a device. \$\endgroup\$
    – Myndex
    Jun 2, 2019 at 18:18
  • \$\begingroup\$ Stumbling upon this years later: Hey to my younger self, Jeeze, stop using words like "absolutely nothing" "never" "always"... and instead "normally not" or "usually"... we live in a universe of non-absolutes... \$\endgroup\$
    – Myndex
    Aug 31 at 12:03

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.