Situation: find color transforms of sufficient CIE-Lab colorspace with closest/widest gamut monitor/... and for Correct Color Management of images computed for devices and human vision
Motivation: to understand some CIE-Lab pictures which can be viewed on closest gamut monitor because CIE-Lab is device independent with lightning conditions) and is perceptually uniform
Use case: Representation of RGB images (etc spectrogram) with correct color management
Images: non-GSDF images where no black/monochrome images, and no graphic elements

Some CIE-Lab spaces where the appropriate question is to consider suitable color transforms for the sufficient CIE-Lab with closest gamut monitor

  • (CIE 1931 possibly not suitable because linear relations)
  • CIE-Lab 1976 possibly not suitable because linear relations, review here interactively
  • CIE-Lab D 50 (many operating systems seem to use this, here wiki)
  • L*A*B discussed mathematically here
  • D65 and D75; normalized to [100, 560] nm here

Possible Workflow for CIE-Lab with encoding non-GSDF images

  1. Preserve linearity of luminance and approximate uniformity of color
  2. rendering intents with re-scaling for perceptually linear color behaviour.
  3. Define such a rendering intent. Etc PLCB is similar to MRC+BPC. TODO
  4. How the mapping can be constrained in the use case? How can you indicate the desired perceptual uniformity of the image through an additional tag?

Generators: Matlab 2016a and Python on Linux Ubuntu 16.04 64 bit and Debian
Monitors: some monitor with correct color management so exclude medical displays of gray-scale, since this approach will replace them
Tests: calibration every 2h proposed to color-radiologists (physical); other tests numeric and analytical
Mathematics: Mathematical Morphology In The CIELAB Space, Web Information Systems Engineering WISE 2014 Workshops on p. 230 by Benatallah and Springer
Physics: ICC review about CIE-Lab

  • This is not exactly a graphics design question. Without a problem to solve I can not answer what to use. – joojaa Aug 3 '16 at 10:05
  • I am sorry, but I don't understand what your question is. – Vincent Aug 3 '16 at 12:35
  • 1
    I think if you take a breath and not dump so much technical spec and write a question worth reading with not so much leading you could maybe ask this on CG.SE youd get much more eyes that can answer but you must be much smarter about asking the question. Right now your asking too much in one go. In essenceyour asking us to do your work for you. Sorry no can do. You can hire me as a consultant but your not going to like my billing. – joojaa Aug 4 '16 at 9:02
  • @Masi how about linear rendering. In either case this is a science question not a graphic design question. – joojaa Aug 4 '16 at 9:27

I am not an expert but, CIE Lab is CIE Lab. What are is evolution and re interpretation of the information.

That is why you have CIE RGB (1920) CIE Luv* (1976) for example. But if you need CIE Lab use CIE Lab.

In reality you are bound to the aplication which you are using to analize the data. For example Photoshop do not let you choose any internal transformation matrix to swich from. (But I do not understand why they use the D50 because a more standarized white point is D65)

The inclusion of ambient light, or lighting conditions is also relative to the size of the screen vs the angle of vision. So it is important that you define a more specific scope.

But to tell you the truth. This is some topics that rarely Graphic designers think of. (shame face) Xo)

Edited acording to the coment below

The D65 is the color temperature for the sRGB standard. http://www.color.org/chardata/rgb/srgb.xalter Most computers are recomended to use it as a base white point. It is not perfect, but the sRGB standard is the most widley used standard across diferent devices.

Some more "specialized" models, like Adobe RGB has diferent white point standards. For example the D50. To tell you the truth I am not sure why. (There is a chance that the reason is because normal papers has a tendency to warmer colors, because oxidation, or other historical reasons, like the usage of tungsten light) But thoose transformations need to be taken in acount by the internal conversion matrix of the aplication, not the Operating System.

Some insights on white

White is a totally relative idea. It is based on the age of our sun, its intrinsec composition of elements, its mass, the absorption of the wave leinght of the Earth'ss atmosphere across the history, and the evolutionary adaptation of our species to that light... = White.

But that is our reference light. But it is also influenced by the ambient atmospheric blue cast on a middle of a sunny day. Take a look at this: https://photo.stackexchange.com/questions/71893/what-is-the-right-displayed-color

  • Great addition! I defined the use case much more clearly now. No need to follow the grayscale standards of radiological images now because the correct color management will replace or has done so already the grayscale radiological workflow, because I have seen color displays being used with radiological workflow in some advanced places. – Léo Léopold Hertz 준영 Aug 4 '16 at 8:46

You use colorspaces to achieve end results. Thus choice of space depends on what you want to do! For a graphics designer it is enough to know what the output is and act accordingly. For this purpose a designer would most likely use either a suitable CMYK/RGB standard space or a more device targetted space. Because that gives the designer control over the output medium, a print or monitor display.

In between it might be useful to use Lab or other spaces to adjust colors more easily. Possibly as device neutral source (for print simulation for example), but in the end that is less relevant as the image needs to eventually display somehow.

There is one big problem with Lab space though. It can make colors that can not be displayed in any system. So you would end up with needing to trust your profile conversion. And in the end your still bound by the colorspace of your output device.

So while Lab is good for certain computation and numeric measurement, you are still bound by your output devices. This is why we use either RGB or CMYK mostly, most of the time we prepare images for consumption. Lab still is not optimal for color interpolation or even light simulation. So at the other end of the spectrum (pun intended) you would need to store actual table of contribution of each spectrum component.

Now the radiologist is still bound by the device he looks through so if you intend the radiologist to benefit from working in Lab space then you need to be calibrating or profiling their monitors/your printers. If all you radiologists have periodically profiled wide gamut monitors then you might benefit from storing and not just manipulating the pictures in Lab space.

Now you would need to define which is more desireable, as wide conversion from your Lab space to your output device or that the devices display as closely the same result each time. And how to handle the colors that lie outside this space. By this tine you have allready dug yourself pretty deep into the near infinite hole of color management.

The reason why designers usually choose sRGB for monitors is because it is most likely to be attainable, that is for most clients even that is beyond their capability. In a closed system calibration is easier to anticipate than maximizing the gamut per device as this would mean monitor A could be better than monitor B and there is not much you can do about this (even if they are the same brand and model), in either case different every time. Even this said you would need periodic recalibration/profiling of monitors or they drift.

And then we have the white/black point and conversion intent discussion. Which I see quite pemature untill you know what problem you are trying to solve.

Personally I would say its safer to work in a device space. This minimizes the risk of puting important data off gamut for the particular device and by misconfiguration in the color matching engine clip the data out. It also ensures that your data is more safe overall if you just ignore the colormatcher alltogether and worry about local contrast. But then i do not know how much of the used hardware you can specify and maintain control over.

  • Great answer! I defined my use case much more clearly in the body. It is correct color management so you do not need to deal with grayscale radiological standards, since the correct color management will replace the use case. – Léo Léopold Hertz 준영 Aug 4 '16 at 8:45

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