Topic: A Hue challenge - the answer, but what should we do?

[themightyzog]

I'm starting a new thread because the answer to the problem raises further questions that I would appreciate some discussion on and therefore hope that those that have already read the 1st thread will read this one.

I had a reply, on another Forum, which said:- "I had a look at sRGB and AdobeRGB spaces compared to my monitor's colour profile, the really saturated corners of red and green are the bits that poke out into the out of gamut areas. Possibly explains your inability to see changes deep into the pure red and greens."

Here are two bands of colour change, each by 6 degrees of Hue. The top one easily shows 6 hues and the bottom one 7. Changing to RGB, instead of this sRGB image, marginally makes it easier to see the green ones.

However, can anybody tell my why assigning no profile does not make it any easier - does this mean that displays are not very good at green?

It appears from my tests that the area 108-120 degrees of Hue are difficult to differentiate, but nature is full of shades of green, which I assume we can see, but screens (and printers?) can not show.

So does this mean that we need to apply the Man from Mars (or similar explosion) technique to this green area in order to display to the viewer what we photographed?

[Greg Groess]

How do they print??

Just curious...

[Greg Groess]

BTW when you select the text of the posting and include the images so there is a blue highlight over the image you can see some of the green steps...I am betting you are right it's the monitor....

Greg

[jacobolus]

If this is HSL/HSV hue, the reason here is that those models are very simple transformations of RGB, designed to be fast to compute on computers of the 1970s, and have little relation to human visual perception.  CIELAB (L*C*h*) hue is reasonably perceptually uniform, but not ideal, and has some nasty curvature of perceptually constant hues as chroma increases, particularly in the blue/purple region.

To get perceptually uniform hue differences, look at your colors in a color appearance model like CIECAM02 (JCh).  Or in the Munsell color system.  Bruce Lindbloom has an ICC profile which sends CIELAB through a look-up table which warps it to match up to the Munsell renotation data; that should work better than just normal CIELAB for getting perceptually uniform hues.

Short answer: HSL/HSV are stupid color models to use in 2008, now that computers are fast.

You may find this article (I never finished writing) useful: http://en.wikipedia.org/wiki/Munsell_color_system

[jacobolus]

BTW when you select the text of the posting and include the images so there is a blue highlight over the image you can see some of the green steps...I am betting you are right it's the monitor....

I can perceive all of the color steps on a properly calibrated display (if I look very closely; I also have pretty decent young-person’s eyesight).  But no, the reason for the difference between the two hue gradients here is not the display.

[Jean Pierre]

Hi,

How many steps is there in the green stripe?

Jean

[themightyzog]

7
can you see them?

[jacobolus]

Oh, one more thing to mention:

The green ramp you have there has pretty constant value/lightness (Munsell value, L*, J, whatever), while the orange ramp has significantly varying value.  Since value contrast is what humans perceive most strongly, this has the largest impact on which of the two appears to have most contrast.

If you normalize L* (dialing back chroma until the colors are all in sRGB gamut), you get the following image, in which you’ll note the colors have completely identical CIELAB L*C*h* hue as the CIELAB hues in the original image you posted (the orange ramp still has significantly more hue variation than the green one).

Edit to add: the forum here seems to strip out the profile from png files when it resizes them to show small in these messages.  I haven't noticed whether it does the same to jpegs.  Either way, that will make them appear incorrect to software which does proper color management for images (for instance, the Safari browser).  So click that image to see the full-size before judging the amount of visible contrast, if you’re using Safari, or ideally open it up in Photoshop. (Or just take my word for this whole discussion, and stop using HSL for any kind of color correction <wink>.)

[jacobolus]

Of course, Photoshop (inexplicably; tragically) doesn’t really provide any particularly good up-to-date color models to work in.  Here’s an image showing the aforementioned hue shift in the blue/purple region.  These colors all have the same CIELAB L*C*h* hue angle.  Still, CIELAB is significantly better with respect to human perception than RGB/CMY, or their associated HSLs.

[KlausNordby]

It appears from my tests that the area 108-120 degrees of Hue are difficult to differentiate, but nature is full of shades of green, which I assume we can see, but screens (and printers?) can not show.

I think you are wrong right there: for why can't nature be "full of shades of green" which we can hardly discriminate between at all -- even though a spectrophotometer will show different values? Human color vision is, well, "pretty good," but just as we cannot see into, say, the ultraviolet (as bees can) there might very well be hue differences within "the human gamut" which it's either hard or impossible for us to clearly perceive. And I think your 7-step green patch shows exactly that: the hue value differences are there in terms of the mathematics, but (at least to my 50-year old eyes) those green hue differences are extremely subtle. However, if you just press Ctrl-I in PS and invert the image, all the 7 magenta steps will then be very easy to see.  So this is not a "monitor issue" or a "profile issue" -- but a "human visual perception issue."

Also, Jacob is right that the orange vs the green patches are not really comparable in a fair way: there are much larger luminosity value differences in the orange image than in the green image, which we are very sensitive to, much more so than to hue differences.

[jacobolus]

However, if you just press Ctrl-I in PS and invert the image, all the 7 magenta steps will then be very easy to see.

This is a meaningless operation (from a human visual-perceptual perspective) unless performed in a perceptually uniform space, so it doesn’t really prove anything at all.