Pages: [1]   Go Down

Author Topic: Imaginary colors and MacAdam limits  (Read 5604 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Imaginary colors and MacAdam limits
« on: February 05, 2017, 06:06:38 pm »

Imaginary colors are created with RGB colorspaces that have one or more tri-stimulus coordinates that are outside the human color gamut. Colors that are outside the Mac Adam limits but not imaginary are colors that can't exist from reflective surfaces under an illuminant such as D50 or D65 no matter how pure the colors. But they can exist from highly pure light sources such as lasers or some LEDs. By definition, printers are limited to colors within the Mac Adam limits. Getting as close as possible to these limits is the holy grail of printer technology.

Imaginary ones can occur with colorspaces like ProPhoto but colors outside Mac Adam limits can occur in images with smaller RGB spaces such as Adobe RGB or even sRGB.

I have a Matlab program that can find imaginary colors but I'm looking for one that can identify colors exceeding the Mac Adam limits. I'm hoping someone here has run across such a beast as I would sure like to add it to my tool set. 
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #1 on: February 09, 2017, 01:44:53 am »

Found what I needed
"Fast and accurate model for optimal color computation" by K. Masaoka

https://www.researchgate.net/publication/44673524_Fast_and_accurate_model_for_optimal_color_computation
Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 608
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Logged

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2370
    • The Last Word
Re: Imaginary colors and MacAdam limits
« Reply #3 on: February 09, 2017, 06:21:57 pm »

Found what I needed
"Fast and accurate model for optimal color computation" by K. Masaoka

https://www.researchgate.net/publication/44673524_Fast_and_accurate_model_for_optimal_color_computation

Thanks for that, Doug.

OBTW, do you know about optprop?

https://www.mathworks.com/matlabcentral/fileexchange/13788-optprop-a-color-properties-toolbox

A really good tool to have in your toolbox.

Jim

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #4 on: February 12, 2017, 03:49:19 pm »

Thanks for that, Doug.

OBTW, do you know about optprop?

https://www.mathworks.com/matlabcentral/fileexchange/13788-optprop-a-color-properties-toolbox

A really good tool to have in your toolbox.

Jim

Yes, it's a good color tool package however, rather than calculating the Macadam limits, it uses published xyY tables from W&S. There is an error in their transcription on line 155 at Y=60  504nm-577nm transition where x was entered incorrectly as .3703 and should have been .2703, a pretty significant error.

Popped out at me when plotting the list. I checked W&S and that was indeed the case. The file name is: Color\DataFiles\OptimalColoursForIlluminantD65.txt

Unfortunately D50 optimal color lists aren't included. Only D65, A and C.

In any case I'll need to compute them for D50.
« Last Edit: February 12, 2017, 05:19:25 pm by Doug Gray »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #5 on: February 12, 2017, 03:59:42 pm »

See also "A new algorithm for calculating the MacAdam limits for any luminance factor, hue angle and illuminant" by Perales et. al.

Yes, that's referenced in their work. Advantage of the new technique is that the code is in Matlab and they used overlapping spectrum windowing to accomplish Type 1 and Type 2 optimal colors w/o separate computations.

I'm considering using the technique but writing it in C++ to create a high rez 3D lookup table for illuminants of interest. Once created I can then just read them into whatever apps might benefit.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #6 on: February 14, 2017, 02:34:07 pm »

Turned out the Matlab code was easily fast enough and there was no need to go to C++.

I've made a couple of charts of the MacAdam Limits in intervals of 5 from 5 to 95 for both L*, which is the L* portion of L*a*b* as well as Y, which is linear. Here are the charts with the 95 value being the smallest area and each lower (90, 85,...) value producing the next larger areas.

Note that these are the boundaries of reflective colors with D50 and so represent the max a theoretical printer could produce even if it had an infinite set of "perfect" inks. The only printable colors that would come close are yellows because there is a wide range of wavelengths that produce yellow and they also are the largest wavelength range with high luminance.
« Last Edit: February 14, 2017, 02:40:35 pm by Doug Gray »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #7 on: February 14, 2017, 11:49:27 pm »

Obviously ProPhoto RGB has many colors that are imaginary and so can't be created with lasers or anything else. But what about lowly sRGB and Adobe RGB (1998)? Obviously they are real colors since each of the R, G, and B colors have chromaticity coordinates that are reasonably inside the human gamut.

So, for fun I looked at sRGB and Adobe RGB (1998) to see if they contain colors that are not even theoretically printable.

Good news. sRGB does not. A printer with unlimited and spectrally ideal inks could accurately print all of sRGB. In theory of course. They don't currently come all that close.

Bad news for Adobe RGB (1998). Such a printer can never exist. Specifically, red is a problem.  Red RGB values of (240, 0, 0) to (255, 0, 0) are not physically printable.  Ever.

This may seem odd since sRGB and Adobe RGB reds have the exact same chromaticies. The reason Adobe fails is that the red luminance at 255 is much higher than sRGB's luminance.
« Last Edit: February 14, 2017, 11:52:51 pm by Doug Gray »
Logged

Ernst Dinkla

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4005
Re: Imaginary colors and MacAdam limits
« Reply #8 on: February 15, 2017, 05:06:04 am »

Obviously ProPhoto RGB has many colors that are imaginary and so can't be created with lasers or anything else. But what about lowly sRGB and Adobe RGB (1998)? Obviously they are real colors since each of the R, G, and B colors have chromaticity coordinates that are reasonably inside the human gamut.

So, for fun I looked at sRGB and Adobe RGB (1998) to see if they contain colors that are not even theoretically printable.

Good news. sRGB does not. A printer with unlimited and spectrally ideal inks could accurately print all of sRGB. In theory of course. They don't currently come all that close.

Bad news for Adobe RGB (1998). Such a printer can never exist. Specifically, red is a problem.  Red RGB values of (240, 0, 0) to (255, 0, 0) are not physically printable.  Ever.

This may seem odd since sRGB and Adobe RGB reds have the exact same chromaticies. The reason Adobe fails is that the red luminance at 255 is much higher than sRGB's luminance.

Research for a neon light image reproduction?

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
February 2017 update, 700+ inkjet media white spectral plots
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #9 on: February 15, 2017, 01:57:55 pm »

Research for a neon light image reproduction?

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
February 2017 update, 700+ inkjet media white spectral plots

In a way, yes.  I have posted on another thread some of my initial experiments re displaying prints that appear to be sourcing light, not just reflecting it. This is done by illuminating an image with a flood where the unevenness of the reflected light from the flood is corrected. This leaves large portions of the print able to reflect light that appears to have greater luminance than is possible with a purely reflected print.

What this makes possible is displaying prints that have highly luminous sources. For instance light streaming through a window or illuminated signage. The goal is to create a print, that when illuminated, looks somewhat self luminous.

It's also been often noted that an image is expected to look different when viewed on a monitor than when printed absent careful comparison with a proof viewing booth. This is almost entirely because monitors are always viewed in rooms with ambient light, and hence max reflected light in the environs, that is a factor of 2 to 5 times more luminous. The MacAdam limits show that even when luminance of WPs are matched, such as in a viewing booth, monitors still render colors near their primaries that can't be reached by the print. In the case of Adobe RGB, can't be reached by any theoretical printer.

So this was just a side diversion mostly done to improve my intuition while pursuing the goal of super luminous print display.  Also, I just find it a lot of fun. I kind of iterate between the intuitive and analytical. They seem to drive each other and it's been my general, productive, approach in many other, unrelated, endeavors.

Logged

Stephen Ray

  • Full Member
  • ***
  • Offline Offline
  • Posts: 217
Re: Imaginary colors and MacAdam limits
« Reply #10 on: February 15, 2017, 11:52:04 pm »

Quote
displaying prints that appear to be sourcing light, not just reflecting it.

I've heard some famous Las Vegas landscape photographer's gallery guests believe the displayed prints are "back-lit." The prints are usually (but not always) Fujiflex which is a reflective film material. However, the material does a very good job at reflecting its source light which is usually a pin-spot gallery lamp such as a Solux or LED MR16 type. "Glowing crystals" may not be far from the truth. If the photo emulsion has a certain micro-structure that lends itself to kick back luminance, the colors can seem to "pop" more than other dyes or pigments that might have a different structure. The optimal structure shape might be round-ish or flat-ish, who knows? But the features remain, something catches and reflects the light better than others.

Dark gallery walls and / or print frame image surrounds significantly help the illusion.

I'm sure the two light source arrangement of the typical spectrophotometer "flattens " or averages the reflectivity of such a material when creating profiles. Does the polarized version help or hinder the results of such a reflective material?
Logged

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Imaginary colors and MacAdam limits
« Reply #11 on: February 17, 2017, 01:06:06 am »

Obviously ProPhoto RGB has many colors that are imaginary and so can't be created with lasers or anything else. But what about lowly sRGB and Adobe RGB (1998)? Obviously they are real colors since each of the R, G, and B colors have chromaticity coordinates that are reasonably inside the human gamut.


An interesting thing is that while ARGB may seem to be within human gamut, that doesn't necessarily mean that it is fully contained within ProPhoto RGB. See the relationship between ARGB and PPRGB below. The area near the saturated unit stimulus blue region of Adobe RGB requires more than unity stimulus from blue ProPhoto RGB primary, i.e., there might be colors that clip for unit stimulus blue primary of ProPhoto RGB but are within the unit stimulus of Adobe RGB.


For more detail please see here:

https://www.dpreview.com/forums/post/37330104
Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #12 on: February 17, 2017, 02:50:51 am »

An interesting thing is that while ARGB may seem to be within human gamut, that doesn't necessarily mean that it is fully contained within ProPhoto RGB. See the relationship between ARGB and PPRGB below. The area near the saturated unit stimulus blue region of Adobe RGB requires more than unity stimulus from blue ProPhoto RGB primary, i.e., there might be colors that clip for unit stimulus blue primary of ProPhoto RGB but are within the unit stimulus of Adobe RGB.

This is interesting and points out a rather curious difference that exists between CIE and ICC terminology that has caused some variation in how color engines work. It will show up Photoshop by doing this.

Take an image made up of all, byte based RGB triplets, in Adobe RGB (1998). Convert them to ProPhoto RGB in Photoshop. Now convert them back to Adobe RGB (1998). What do you get?

Well, it depends.

Rounding errors of course. But clipping going from Adobe RGB to ProPhoto? Only in one case and that case involves differences that show up in Absolute Colorimetric and only when using Microsoft's ICM engine. Then and only then, you get the clipping you describe because that engine attempts to preserve the D65 white point of Adobe RGB which it cannot for the reasons you describe.  There is not sufficient amplitude to bump up the Z component to the required level demanded by D65.

This, it turns out, was due to an ambiguity caused by a different usage of Relative and Absolute Colorimetry terminology between the CIE and ICC.  ICC addressed this about a decade ago by stating that all conversions of matrix colorspaces must use color whitepoint adaptation transforms when changing whitepoints. Even with Absolute Colorimetric Intent.

Adobe's default color engine, ACE, observes this ICC requirement and produces no clipping.

Logged

Ernst Dinkla

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4005
Re: Imaginary colors and MacAdam limits
« Reply #13 on: February 17, 2017, 05:56:04 am »

In a way, yes.  I have posted on another thread some of my initial experiments re displaying prints that appear to be sourcing light, not just reflecting it. This is done by illuminating an image with a flood where the unevenness of the reflected light from the flood is corrected. This leaves large portions of the print able to reflect light that appears to have greater luminance than is possible with a purely reflected print.

What this makes possible is displaying prints that have highly luminous sources. For instance light streaming through a window or illuminated signage. The goal is to create a print, that when illuminated, looks somewhat self luminous.

It's also been often noted that an image is expected to look different when viewed on a monitor than when printed absent careful comparison with a proof viewing booth. This is almost entirely because monitors are always viewed in rooms with ambient light, and hence max reflected light in the environs, that is a factor of 2 to 5 times more luminous. The MacAdam limits show that even when luminance of WPs are matched, such as in a viewing booth, monitors still render colors near their primaries that can't be reached by the print. In the case of Adobe RGB, can't be reached by any theoretical printer.

So this was just a side diversion mostly done to improve my intuition while pursuing the goal of super luminous print display.  Also, I just find it a lot of fun. I kind of iterate between the intuitive and analytical. They seem to drive each other and it's been my general, productive, approach in many other, unrelated, endeavors.

Not as advice to go that way but fluorescent inkjet inks exist and I think it was Roland that offered them as OEM inks for certain printers.

I think it would be possible to create an embossed paper texture that is near retroreflective. Reflect the light more bundled yet preferably at another angle than the incoming display light has. It would require a precise placing of the light and of the observer. Glass pearl projection screens in a way. Some paper surfaces already aim at an effect like that already.

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
February 2017 update, 700+ inkjet media white spectral plots
Logged

BradFunkhouser

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
Re: Imaginary colors and MacAdam limits
« Reply #14 on: February 17, 2017, 10:37:18 am »

Bad news for Adobe RGB (1998). Such a printer can never exist. Specifically, red is a problem.  Red RGB values of (240, 0, 0) to (255, 0, 0) are not physically printable.  Ever.

This may seem odd since sRGB and Adobe RGB reds have the exact same chromaticies. The reason Adobe fails is that the red luminance at 255 is much higher than sRGB's luminance.

So, there aren't any (non-fluorescent) inks that can actually produce those AdobeRGB reds under D50 light.  Very interesting.  I'm curious what happens when you use Illuminant A (tungsten) instead of D50?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Imaginary colors and MacAdam limits
« Reply #15 on: February 17, 2017, 02:28:54 pm »

So, there aren't any (non-fluorescent) inks that can actually produce those AdobeRGB reds under D50 light.  Very interesting.  I'm curious what happens when you use Illuminant A (tungsten) instead of D50?

I haven't built tables for Ill. A, but a look at the relatively coarse tables in S&W shows it to be close. There's a good chance aRGB would be theoretically within MacAdam at max red.
Logged

joofa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 544
Re: Imaginary colors and MacAdam limits
« Reply #16 on: February 17, 2017, 04:13:27 pm »

Rounding errors of course. But clipping going from Adobe RGB to ProPhoto? Only in one case and that case involves differences that show up in Absolute Colorimetric and only when using Microsoft's ICM engine. Then and only then, you get the clipping you describe because that engine attempts to preserve the D65 white point of Adobe RGB which it cannot for the reasons you describe.  There is not sufficient amplitude to bump up the Z component to the required level demanded by D65.

Yes, I agree with your explanation.

This, it turns out, was due to an ambiguity caused by a different usage of Relative and Absolute Colorimetry terminology between the CIE and ICC.  ICC addressed this about a decade ago by stating that all conversions of matrix colorspaces must use color whitepoint adaptation transforms when changing whitepoints. Even with Absolute Colorimetric Intent.

Adobe's default color engine, ACE, observes this ICC requirement and produces no clipping.

It is nice to know that. Thanks for the info.
Logged
Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins
Pages: [1]   Go Up