Color Management in Raw Processing

(Colin Paul Adams) #42

Because you need a base profile to convert from later

(Glenn Butcher) #43

At the very beginning, when the raw file is opened.

If all that was done there was to look up the camera’s primaries in the big list, it might not be important to distinguish the act of ‘assign’. But, if user-produced camera profiles are considered, assignment becomes a user concern.

(Hermann-Josef) #44


Because the profile information is needed to convert the pixel values to device independent values. These are then converted with the output profile e.g. of the monitor to calculate the values in RGB to be sent to your (calibrated) monitor to see there the correct colours. This is the whole issue of colour management.


(Glenn Butcher) #45

The internal mechanics of conversion, inputprofile -> XYZ -> outputprofile, is a very important concept in its own right. It decouples softwares from having to know the colorspace end-to-end, because there’s a reference space that both sides explicitly understand and convert to/from. That’s one of the things i don’t see clearly in OCIO, but I need to study it further.

I’m trying to avoid these internals in this graphic, want to keep it at an overview level. But once the flow is understood, those internals rise to prominence…

(Glenn Butcher) #46

Really, maybe ‘convert’ should be considered as two steps: 1) transform the image, and 2) tag it with the new colorspace metadata. Which is really ‘convert’, followed by ‘assign’… :smiley:


@Jossie & Colin_Adams

Yes. But I am trying to put myself into the brain of a “normal” user.
Target group analysis…

(Alberto) #48

hi @Claes,

let’s see if this helps: is 10 a good price for an espresso? well, you can’t really answer unless you first assign a currency tag to that 10, right? and after doing that, how do you judge whether the price is right? one possibility is to convert to your local currency, and then check whether it is in your price range. so, for example if I’m in Stockholm and get charged 10, I first assign the ‘SEK’ profile to the data, then I convert it to the ‘EUR’ profile, and, after I get about 0.98, I know the price is fair :slight_smile:


Ciao! Not bad at all :slight_smile:
The problem is, however, how to translate that into
development of images?

As I see it, a target group analysis really is called for.
Is the target’s knowledge of the subject == Elle’s?
(If so, you do not need an explanation at all!)

If the target group knows nothing [which in my opinion
ought to be the case here]: how to describe it in an understandable way?
Easy way out: if it becomes far too complicated to explain, better skip the explanation?


Making me think of the game show.

I use Inkscape for all figures and charts because I can customize it however I please and don’t have to bend over backwards to get a flowchart, mind map, concept map, etc., app to follow my exact specs.

(Glenn Butcher) #51

Well, I’m using my own recent experience. About two years ago, ‘color management’ for me was insuring I put the crayon back in the box in the same place where I got it. Two years later, I’m still struggling with the things like the implications of metamerism, but I’m a lot more savvy about the mechanical parts of the process. And, if I’d have had an exposure to the mechanical overview of things earlier, my wading through the technical details would have been less painful.

I look at the cartoon as the outer skin of the onion. Study it, then peel it back and explore the next layer, and so on…

(Hevii Guy) #52

Trying to… catch… my… breath and… stop laughing… so that I can post this reply :rofl::joy::rofl:


Must. Make. Sure. That. My. Crayon. Is. Colour. Managed. OH-WAIT.

(Glenn Butcher) #54

Oh, and one year prior to that, ‘bit-depth’ was how far the drunk at the bar sunk his teeth in to my leg…

(Glenn Butcher) #55

I hesitated to use the crayon analogy, there’s a poor fellow over a dpreview who has caught holy you-know-what over crayon analogies and color spaces. He’s just trying to explain this stuff in a way he figures regular folk will understand. Not sitting well with the polymaths…

Analogy sometimes brings out the worst in people… :fire:

Edit: Analogy does beg imprecision, no doubt. And sometimes, one person’s analogy is just another’s irritating set of words. What it comes down to is that posting in forums is not the same as talking face-to-face, or any of the other ways we communicate. Gotta know thy audience.


The assign camera profile stage in raw converters like darktable happens after the working profile conversion, I think (actually the manual says it’s a conversion step, not an assignment step). Here’s a screenshot of an example, basic pixelpipe, where the modules at the bottom happen first:


Input color profile gets assigned after WB, demosaicing and exposure modules, etc.Here’s how the manual describes what happens with colour space transforms.


I think it doesn’t matter whether the assignment is before or after those steps since they are profile agnostic so to speak.

(Glenn Butcher) #58

I don’t think the conversion to the working profile e.g., ProPhoto, can happen before the camera profile is assigned.

No, I’m sure of it. I can’t think of how cmsTransform would work against the data without first assigning the corresponding camera profile to the camera raw.


Correct! We call it target group analysis :slight_smile:

(Glenn Butcher) #60

That is true, the operations themselves don’t particularly care about the particular color space. However, some operations behave better on the linear data, and the exposure right after demosaic starts to perturb that. Also, I recall in @Elle’s writings some description of the benefit (or avoidance of detriment, can’t recall which) of editing in a well-behaved working profile, which would mean that it would be better to do that conversion earlier…

(Glenn Butcher) #61

And my first analytical assertion is our target audience doesn’t like analogy… :smile: