Equal Luminosity

[holyeyeolo]

Hi,

I'm trying to find a way to choose equal Luminosity colors. From what I understand Luminosity is the PERCEIVED Brightness by a human viewer, and is thus different from the Lightness of sRGB workspaces. My question is, is it also different from the Lightness of LAB? Explanation of luminosity in Photoshop  and Re: Luminosity/Color blend modes - what do they really do?  give different answers. If I open a file in Color Mode - Lab Color, then in the color picker I pick 2 different colors in the Lab color selection WITHOUT altering the L, would that give me two colors with similar Luminosity? Two colors with similar perceived Brightness to the standard human eye?

Also, is desaturating them a good way to check that? Wouldn't that mean that desaturatin a color in Lab mode different from desaturating a color in RGB mode?

Thank you

[c.pfaffenbichler]

Also, is desaturating them a good way to check that?

Well, what do you mean by that?

Desaturating with the Hue/Saturation Adjustment in RGB averages the three Channels, so that would not give you a proper representation of Luminance as evidenced in Lab.

[holyeyeolo]

Not in RGB, in Lab mode. If I desaturate two colors In RGB color mode and they give the same gray, then they both have the same Brightness. If I desaturate two colors in Lab color mode and they give me the same gray, then they both have the same Luminance. If I desaturate two colors in RGB mode that have the same L in the Lab color selector, then they won't necessarily give me the same Brightness. Is that all correct? 

[G_Hoffmann]

Nobody in color science knows, how Photoshop´s luminosity is defined.

Therefore I don't refer to this expression.

R.W.G.Hunt says in his great book Measuring Colour:

"3.2. The Y tristimulus value

If the Y tristimulus value is evaluated on an absolute basis as Y (with index L),

in candelas per square metre, for example, it represents the luminance of the colour.

This provides the basis for a correlation with the perceptual attribute of brightness.

As has already been explained, the correlation is complicated by the effect of the

viewing condiditions, by the non-linear relationship between brightness and

luminance, and by the partial dependance of brightness and colourfulness."

(bold by me)

Y is a linear combination of R, G and B in any linear RGB color space, that is to say

without any gamma encoding.

L* in CIE 1976 (CIELab) is called CIE 1976 lightness or simply Lab lightness.

The relation between Y and L* is nonlinear but monotonic: a series of patches

with increasing values Y has as well increasing values L* (and patches with identical

values Y have as well identical values L*).

But alas, such a series doesn't show monotonously increasing brightness (visual

appearance) if the colors differ considerably by saturation.

Open page 10 in this doc

http://docs-hoffmann.de/gray10012001.pdf

in Photoshop in mode LAB (!).

Each image is built in index color with 256 colors in the palette. The palette diagram

shows these colors ordered by values of Y or  L* (which delivers the same order).

That is, the Lab values are monotonously increasing, but the visual brightness seems

to jump. This is mainly a result of the so-called Helmholtz-Kohlrausch effect: more

saturated colors appear brighter.

(The order is almost monotonous, there are small deviations because of various

roundings in the calculations, please don't worry.)

The result: the bold text as cited above – there is no simple recipy for colors with

equal perceived (visual) brightness.

Best regards  --Gernot Hoffmann

P.S.:

This doc

http://docs-hoffmann.de/palette30082003.pdf

shows as well, that the attempt to create "balanced palettes" with colors of equal

brightness based on equal values L* failed partly.

[holyeyeolo]

"the correlation is complicated by the effect of the

viewing condiditions"

The viewing conditions are engaged by industry professionals by functions of Color Tolerance such as CMC/CIE94 (https://www.xrite.com/-/media/xrite/files/whitepaper_pdfs/l10-001_a_guide_to_understanding_color_com...  p16-21) if I understood correctly. There's no mention of Photoshop using such thing (CIELAB color space - Wikipedia), so I'll settle for ignoring those for now.

"partial dependance of brightness and colourfulness."

I am aware of the Helholtz-Kohlrauch effect. I'm also aware that CIELAB doesn't take it into account. But I just need the limits of what the software can give me, more than that is up to my own eyes and taste.

"non-linear relationship between brightness and

luminance"

That's the point here. If CIELAB inaccurately solves this relationship without considering the other two problems, then that's all I'm asking for. Does it?

Again, I just want it to lead me in the right direction, I'm not likely to take the viewing conditions into account any time soon, and the Helmholtz-Kohlrauch effect is left to my eyes. I still wonder how Monet for instance picked colors of similar Luminosity by eye alone. But if he did it, then it's possible. If I may ask, do you know of a book, perhaps more artist oriented and a bit smaller than R.W.G Hunt's book (unless you tell me that's the right one) that could teach me some good techniques on these matters?

[D Fosse]

Actually I think Luminosity in Photoshop is identical to the L channel in Lab.

That makes sense, since Lab is always present in the background as a Profile Connection Space (PCS) in all color management operations. There's always a Lab value behind any RGB value, and it's always readily available.

This explains itself:

[holyeyeolo]

That's what I wanted to verify. Since in this thread Re: Luminosity/Color blend modes - what do they really do? (from 2008 mind you) see comments 2 and 6, the opposite is stated - Luminosity and L* of Lab are different. While in your comment from 2016 Explanation of luminosity in Photoshop  they are the same. So as of 7/10/18 Luminosity and Lightness of Lab in Photoshop are the same. Is that correct?

I got really confused because in other softwares the Luminosity functions have no connection to the meaning of the word. They are like HLS workspace or something... I downloaded Photoshop Trial ONLY to check CIELAB out.

[D Fosse]

You can verify for yourself. Open two copies of the same image. Convert one to Lab and desaturate.

In the RGB version, you need to keep gamma constant in order to not get false results. Grayscale is displayed according to your working gray - and what you want to do is display a luminosity mask which is grayscale.

So to keep this constant, set your working gray to gamma 2.2, and convert the file to Adobe RGB which also has a straight gamma 2.2 tone response curve (sRGB has a custom TRC, and ProPhoto 1.8).

With this set, make a luminosity selection, ctrl + 2, and convert to mask by adding an adjustment layer. Now alt-click the mask icon to display it as a grayscale image.

These two should now match, if my theory holds. I haven't done extensive testing, but so far it seems to hold up.

[holyeyeolo]

Sorry, I'm not used to Photoshop so I might have done something wrong. I did everything exactly except ctrl + 2 doesn't work so I ctrl + clicked the composite RGB channel. What type of adjustment layer? I tried a black/white one, and I tried to simply make a new layer and then click on the mask icon to create a mask on it. Then alt clicked the mask icon. both option yield different results from the desaturated Lab one. But to be honest, I've no idea what's the theory behind the "working gray", so I don't understand what I just did.

[D Fosse]

Sorry, that's ctrl+alt+2.

The type of adjustment layer doesn't matter, you just want the mask which shows the Luminosity transfer function. You can't have a mask without a layer, any layer.

The working gray part is extremely important. The mask is a grayscale image, and grayscale images are displayed with your working gray profile assigned. You need to have an RGB document profile and a mask grayscale profile, with matching tone response curves (gamma). Otherwise they won't be comparable. The easiest way to ensure this is to use Adobe RGB and Gray Gamma 2.2, which have identical curves.

[holyeyeolo]

Alright, I used Adobe RGB (1998) D65 WP Gamma 2.2 last time because of the Gamma 2.2 bit. Now I changed to the regular Adobe RGB (1998) and then repeated the process. The two pictures are now indeed exactly the same to my eyes. I'll have to read on that "matching tone response curves (gamma)" thing tomorrow, from just a bit of reading it seems to also be connected to the human perception of brightness somehow.

Thank you very much D Fosse and G.Hoffmann!

[D Fosse]

holyeyeolo  wrote

Alright, I used Adobe RGB (1998) D65 WP Gamma 2.2

That's some manufacturer monitor profile. So no, that's not what you need, it will be highly irregular.

[G_Hoffmann]

Continued about colors with equal brightness:

Munsell colors were made by the inventor for equal visual brightness for the same "Value",

which is the numerical Munsell parameter for lightness / luminance / brightness.

http://docs-hoffmann.de/munsell15052009.pdf

It's surprising that the Lab lightness is almost constant for the same Value. The Lab numbers

are shown on reference pages for the color tables.

If sorting by Lab doesn't show similarly sorted brightness, then sorting by Value wouldn't be

very different. Still an enigma...

Best regards --Gernot Hoffmann