I'm back down the colour management rabbit hole . . . Canon RAW Gamut & Colour Space (6D & 5D3)

So, someone said something to me yesterday that has me scratching my head and looking for documentation. I'm having a terrible time turning documentation up in searching Google.

My understanding, for the last several years, is that when a photograph is captured by the sensor it is 14bit and has no colour space, but when written to the RAW file it is 14bit LAB (with the sRGB or aRGB settings just affecting LCD preview and JPGs).

From what I read yesterday (a DP Review thread from 2004), it may actually be 14bit sYCC (YCbCr colour space). I had never heard of this before, but apparently it is an extended format RGB colour space. I don't know if Canon moved to using LAB more recently, or if they've used sYCC since 2003 (when the DP thread indicated the colour space information was from).

What was said to me yesterday was that my camera RAW files were RGB. My "inside voice" said, "No. They're LAB." But, if they're sYCC they are probably technically RGB . . . in which case it is a good thing I didn't use my "outside voice" to express what I thought was fact.

I'm regularly wrong about things, and I'm perfectly content with that, but I hate to discover that I'm wrong after emphatically arguing my erroneous understanding of the facts with someone. I just wind up feeling like an opinionated and uneducated jerk.

Whether this person was right or wrong doesn't matter to me. Whether right or wrong I won't be discussing it the person in question. I respect the person, and I have no desire to engage in a pissing match, or let the person know that I even questioned his/her knowledge on the matter. It just isn't important. Life is too short to complicate relationships with people I enjoy and respect.

Interestingly, it was my first Photoshop instructor who told me my RAW files are LAB. It is my second one who told me my RAW files are RGB.

What is important to me is that I may have developed my own personal colour management system based on an error in my understanding of the original RAW file colour space that I'm starting with.

I'm looking at the graphic posted here: http://sonyalphalab.co​m …and-which-should-you-use/.

My understanding is that LAB covers the coloured area of the diagram (the gamut of human vision). As a result, I edit in 16 bit ProPhoto RGB. I want to cover as much of the original colour space of the RAW files as possible when editing them. ProPhoto RGB includes theoretical colours, and it doesn't cover all the visible gamut, but it is the closest I can get without editing in LAB. I'll export my edited photos to 8bit sRGB. Only in very rare instances will I export to aRGB, and in even rarer circumstances to CYMK. I'll sometimes use LAB as an intermediary between ProPhoto RGB and CYMK.

In looking once again at the diagram, I'm wondering if the sYYC colour space is closely represented by the "Wide Gamut RGB" triangle, or if it covers a different region of the gamut of human vision. If it is the wide gamut RGB triangle, then I think it still makes sense for me to edit in ProPhoto RGB (unless I want to move to editing in LAB, which I don't think I want to do . . . should I? . . . maybe I should try).

I use a 6D and a 5D3. Can anyone direct me to any official documentation to confirm the colour space of my RAW files? If no official documentation is known of, I'll settle for an educated hypothesis and try to do my best with that. While I'm at it, maybe I should also confirm that my RAW files are actually 14bit.

In addition to being a photographer, I'm a specialist in digital cartography. The two fields are incredibly closely related in the digital world, and understanding colour and colour management is very, very important to me ("anal retentive" might be a good description of me in this context): whether I'm taking a photograph from capture to print or a digital map from the planning stages to plotting. I thought I had things figured out for my RAW files, but now I'm questioning things and trying to sort through this newly introduced confusion and the seeming lack of available technical information is very frustrating.

I would be greatly appreciative if anyone can help to clarify things for me.

My view on this.

In both RGB and Lab color systems, a pixel's color is determined by three parameters.

In the RAW file, for each pixel, there is only the information about the amount of the single (filtered) color component, either Red, Green or Blue. It's only after demosaicing, performed by the raw converter, that a pixel is assigned the complete RGB triplet, which determines its final color.

Well, let's hear the gurus!

Guru # 1 might say...I work in LR/PS CC and use ProPhoto for my PP, having learned from the majordomo of color theory, Andrew Rodney http://digitaldog.net/
After PP is accomplished, I would consider the end use of my work...sRGB for internet taking browser profiles into consideration,
or, softprooofing for whatever printer profiles are required. One must consider that every monitor/every browser/every printer coughs out different results.

In the end, the only thing that matters is the "eye"...how it perceives your object d'art.

RAW does not inherently 'have a color space'.

1. 14-bit RAW merely records the electrical response of each single color sensel on the sensor;
   
2. then, later groupings of two Green, one Red, and one Blue sensel are used collectively to later determine a pixel value in an R-G-B color space,
   first a very large color space within Lightroom during postprocessing
   
3. and then a limited 8-bit color space when the JPG is output.
   

I don't mean any offence, but ... you know there are thousands of people out there creating beautiful images while you're worrying about this stuff, right?

Thank you. It seems so obvious now that it has been laid out the way it has.

So, RGB is absolutely correct.

I wonder if instructor #1 was referencing the gamut of human vision when he referenced LAB.

So, what is important is the gamut covered by my cameras' sensors, and determining what colour space completely encompasses that gamut: without going too far beyond it.

I'll see if a Google search can provide the information with regards to the cameras' sensors.

Damo77 wrote in post #18023974
I don't mean any offence, but ... you know there are thousands of people out there creating beautiful images while you're worrying about this stuff, right?

Yes. But, most of those thousands of people don't have the same relationship with data that I do. They don't deal with the various remote sensing technologies (including LiDAR, satellite DEMs, satellite imagery, and aerial imagery). They don't process survey data (current or historical), generate 3D surfaces, or generate digital and hard copy map products. They don't spend their lives worrying about data integrity and quality to anywhere near the same degree that I do. In legal land surveys, data must be clearly understood and properly managed.

I can't turn it off. It is part of what makes me so good at what I do. It also makes me a bit of an anal retentive freak in this department.

Edit: I'll add that if I make an error with a dataset and no one catches it, it could cost millions to correct and potentially years of litigation. As it presently stands, I've only made one costly mistake in twenty years. Fortunately it only cost around 12k, and it didn't result in litigation of any sort. I somehow mirrored the coordinates for a proposed pipeline when I created a GPX file (for a helicopter scouting mission) on a weird angle bisecting the equator and the proposed pipeline wound up in Brazil instead of Alberta. To this day I have no idea what I did, and I can't believe that no one who proofed the file caught it before the file was loaded into the helo's GPS system.

Damo77 wrote in post #18023974
I don't mean any offence, but ... you know there are thousands of people out there creating beautiful images while you're worrying about this stuff, right?

i was going to say something similar. I first started color correcting images (professionally) and then converting RGB > CMYK back in the mid 90s with Photoshop 4, we used Matchprints® before we went to film. It was quite a process to insure accurate color for apparel manufacturers catalogs that were sent to retailers. We had to be perfect. However, we only worried about what we needed to know, and there was plenty we didn't need to know.

things are pretty darn simple now.

neacail wrote in post #18024027
Yes. But, most of those thousands of people don't have the same relationship with data that I do. They don't deal with the various remote sensing technologies (including LiDAR, satellite DEMs, satellite imagery, and aerial imagery). They don't process survey data (current or historical), generate 3D surfaces, or generate digital and hard copy map products. They don't spend their lives worrying about data integrity and quality to anywhere near the same degree that I do. In legal land surveys, data must be clearly understood and properly managed.

I can't turn it off. It is part of what makes me so good at what I do. It also makes me a bit of an anal retentive freak in this department.

I see your point of view too.

I'm not saying you should, but i try to redirect some of that energy to other things when I feel I'm getting too far down the rabbit hole. That is unless i seriously have nothing better to do.

The way I look at it is that RAW has no color space.

We are constantly bombarded with radio waves from space, the background noise of the universe. If we hook up just the right equipment we can hear those waves, but without them, we cant. Once those satellite receivers are hooked up and the computer starts recording and processing, they can be moved into another format that is easily processed as audio. Same with light, sensors and recording media. It doesn't exist as an image/sound until is is processed into a format that can be seen/heard.

the image on your screen, and the sound coming from the speaker are the first iterations of the source that can be processed by us humans. It is at that point (or slightly before) that they are assigned something approaching a color space or audio format.

film, monitors, and output devices have gamuts and color spaces, RAW does not.

dang, that was more than i intended to type out, i need to redirect my energy to this cold beer sitting beside me.

I'm suddenly happier I didn't complete my geology major.

neacail wrote in post #18024027
Thank you. It seems so obvious now that it has been laid out the way it has.

So, RGB is absolutely correct.

No the RAW file contains ZERO specific colour information in the data. All of the colour information is infered by the demosaicer when it processes the image. The demosaicing algorithm has data about the spectral responce of each colour of filter in the Bayer CFA. Since it also knows that those filters are arranged as alternating Red, Green and Green, Blue rows. It is possible to then calculate a set of RGB, CMYK, or even LAB values for each pixel location from the knowlege of the filter responces and the recorded value of each sensel, which in Canon bodies since the 40D have been recorded as 14 bit values. In older cameras it was 12 bit. It is the demosaicing software that decides what sort of colour system you will end up with, and then specific colour space within the colour system.

Using a program like RAW Digger it is possible to look at the RAW data values for each sensel. It will become very clear if the RAW file uses 12 or 14 bits from the maximum sensel value. 12 bit will have a max sensel value of 4096 while 14 bit can reach 16384. The program will also show you true RAW histograms based on the RGGB CFA arrangement of the sensor.

Alan

Am I correct in assuming that the OP is not referencing photography as we know it?

I like to think of RAW as RGBG. I think some sensors have done other things. Foveon with RGB pixels. Sony with RGBE(merald) cybershots.

Benitoite wrote in post #18024270
I like to think of RAW as RGBG. I think some sensors have done other things. Foveon with RGB pixels. Sony with RGBE(merald) cybershots.

That is exactly where my understanding has gone with this RGBG (or simply RGB, so my mildly dyslexic brain doesn't jumble the position of the first G).

I need to mull this over some more, but things are coming together.

chauncey wrote in post #18024215
Am I correct in assuming that the OP is not referencing photography as we know it?

I really am down the rabbit hole on this one. But, my brain is working through it.

Bigal007 has this one right. I think it was Andrew Rodney that said that RAW files don't even have a gamut. A gamut is an artificial limit on a color space. The RGBG pixel's color filters don't have that limit on response.

But the simple answer to the question that I didn't see the OP ask, is: use the post-processing gamut that just contains the color information you want to retain. In many cases, Adobe RGB will give you the extent of coverage you need, while not posterizing things like the sky. If you are shooting flowers, or highly-saturated dyes or paints, you might need a wider gamut that won't clip important colors.

skid00skid00 wrote in post #18024361
Bigal007 has this one right. I think it was Andrew Rodney that said that RAW files don't even have a gamut. A gamut is an artificial limit on a color space. The RGBG pixel's color filters don't have that limit on response.

But the simple answer to the question that I didn't see the OP ask, is: use the post-processing gamut that just contains the color information you want to retain. In many cases, Adobe RGB will give you the extent of coverage you need, while not posterizing things like the sky. If you are shooting flowers, or highly-saturated dyes or paints, you might need a wider gamut that won't clip important colors.

Just remember that posterisation has as much to do with bit depth as it does with the gamut of the chosen RGB colourspace. Which is why you should be doing as much editing as possible in as many bits as possible. When I have to edit in RGB, rather than working on a RAW file, I try to keep everything in 16 bit, where you should get a much smaller step between colours in ProPhotoRGB than you do in even 8 bit sRGB. Also I would always do the colourspace conversion first, then throw away the extra bits of information.

Alan