RGB? Lab? HIS? HSV?

[Hening Bettermann]

This Wikipedia article
http://en.wikipedia.org/wiki/HSL_and_HSV

is rather critical to HSV/HSL/HIS color models for editing, and advocates Lab. In contrast, Tindemans, in a comment to his Tonability plug-in, says

"Using complete hue and saturation locking in RGB mode is equivalent to applying a brightness curve in the HSB representation. HSB brightness appears to be a surprisingly good measure for perceived illumination."
http://simon.tindemans.eu/tools/tonability

I also remember that Andrew Rodney states that Lab is not perceptually uniform.

PhotoLine offers editing in HSV and HIS, in addition to RGB and Lab. So I did a little test.
The test image is too dark. I applied a Spline curve of x=50%, y=70% in an adjustment layer.

For the Lab, HIS and HSV models, PhotoLine allows the curves applied to either just the Lightness/Brightness ("grau" in my image titles) or to all channels, and I tried both.
Here are the results as screen shots from my not-hardware-calibrated MacBook Pro Retina screen. Png's converted to jpeg's of 70% quality.

My choice is HSV, either just the V ("grau") or all channels.

[Hening Bettermann]

a little text to justify the post for the attached images

[Hening Bettermann]

a little more text to justify even more attachments

[AlterEgo]

Lab is not perceptually uniform.

http://brucelindbloom.com/UPLab.html

[torger]

The various "perceptually uniform" CIE color spaces are quite far from being that as they have requirements on them to be linear and reversible etc so they can be useful in various color transformations. They are as perceptually uniform they can be with the given mathematical constraints. CIE in general prioritizes mathematical usefulness over perceptual accuracy.

The nonuniformity is shown as 1) despite constant hue angle (Lab in angular form = LCh) the perceptual hue changes with increased saturation (the well-known "blue becomes purple" problem), 2) moving 1 unit distance in any direction is not really perceptually uniform, eg moving 1 unit near the whitepoint is seen as a larger color difference than moving one unit from a very saturated position. That's why the complex CIEDE2000 algorithm is preferred for color difference rather than simple distance in Lab space if perceptual accuracy is important.

RGB-based spaces, which unlike Lab, Luv Jab etc is not unbounded but constrained by the chosen RGB primaries, do not have the hue drift issue. However they are still nonuniform in terms of color difference, as far as I know more so than Lab.

When it comes to the lightness axis I know less about nonuniformity, but I guess there are issues there too, but in that case I'm quite sure Lab is more uniform than the typical RGB-based space. Please correct me if wrong.

Say if you want to increase saturation in angluar Lab (=LCh) by moving the chroma slider, you may get a color shift. Moving the saturation slider for HSL or HSV is then much better as it won't change hue.

Personally I think Lab is useful for scientific applications and profiling etc as it's unbounded. The perceptual uniformity issues are rarely a problem in those applications as you're rarely into making subjective saturation adjustments. When it comes to image editing I think the RGB spaces are generally better suited. A problem with RGB is color shift when you change contrast though. There is no single perfect color space that can do it all out there.

[Hening Bettermann]

Thank you for your comments.
What strikes me in my little test is the desaturation that occurs in all but the HSV space, also in standard RGB. So it seems Tindemanns is right.

[digitaldog]

Personally I think Lab is useful for scientific applications and profiling etc as it's unbounded. The perceptual uniformity issues are rarely a problem in those applications as you're rarely into making subjective saturation adjustments. When it comes to image editing I think the RGB spaces are generally better suited.

Couldn’t agree more. And if Bruce Fraser were here, he'd say the same and did. Many, many years ago on on the CS list:
Bruce Fraser wrote:

Let me make it clear that I'm not adamantly opposed to Lab workflows. If
they work for you, that's great, and you should continue to use them.

My concern is that Lab has been oversold, and that naive users attribute to
it an objective correctness that it does not deserve.

The only real advantage Lab offers over tagged RGB is that you don't need to
send a profile with the image. (You do, however, need to know whether it's
D50 or D65 or some other illuminant, and you need to realise that Lab (LH)
isn't the same thing as Lab.) In some workflows, that may be a key
advantage. In many, though, it's a wash.

One thing is certain. When you work in tagged high-bit RGB, you know that
you're working with all the data your capture device could produce. When you
work in Lab, you know that you've already discarded some of that data.