Pages: [1]   Go Down

Author Topic: Test Image for Checking Wide Gamut Monitors  (Read 8449 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Test Image for Checking Wide Gamut Monitors
« on: August 28, 2016, 01:36:30 am »

This image, in ProPhoto RGB, is composed of two columns. The left column of rectangles will appear to have different colors on the left and right sides of each rectangle with the right hand side being more saturated.

However, on a wide gamut monitor the right hand column of rectangles should appear just one color each and the same as the right side of each rectangle in the left column.

A regular, sRGB monitor will display the right column of rectangles very similarly to those on the left.

The left side of each column of rectangles is actually composed of different, alternating colors that, when combined visually merge. If you have a high (4k) monitor you won't see the distinct lines but may if you zoom in. If you see distinct lines on a low res monitor back off until they disappear.


Logged

rasworth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
Re: Test Image for Checking Wide Gamut Monitors
« Reply #1 on: August 28, 2016, 11:06:21 am »

Doug,

Nice tool, works as advertised on my SW2700PT.  Any objection if I save and use for evaluating others' monitors?

How about an image that verifies sRGB coverage?

Richard Southworth
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #2 on: August 28, 2016, 12:30:53 pm »

Doug,

Nice tool, works as advertised on my SW2700PT.  Any objection if I save and use for evaluating others' monitors?

How about an image that verifies sRGB coverage?

Richard Southworth

You, and others, are welcome to use it but it's just an initial cut. I made it in Matlab using sRGB an Adobe RGB gamuts then matching the XYZ averages of the stripes to produce the same values as those in the solid sRGB gamut edge colors.

It's a first cut. I'm looking at expanding it to show larger differences and using deltaE2000 which is perceptually better at identifying color differences.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #3 on: August 28, 2016, 10:07:06 pm »

I have improved the color separation between sRGB and Wide gamut monitors by using a repeating, 4 pixel pattern. It contains 3 pixels at the Adobe RGB gamut boundary and a fourth pixel whose color has been adjusted so that the average X,Y, and Z CIE colors are the same as the X,Y, and Z colors on the right side of each rectangular patch. By using 3 of the 4 pixels the effect of clipping is much more pronounced.

I have provided two tif files. One in Adobe RGB (1998) color space and the other is the image converted to sRGB. Both files should appear the same on monitors that are close to or profile to emulate sRGB monitors. On a wide gamut monitor one of the tif files will appear to be 6 color patches of a solid color and the same on the left and right.

It's important to view these at 100% (one to one image to display pixel). For lower resolution monitors you should also move far enough away that the 4 pixel pattern disappears.
Logged

Lundberg02

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 379
Re: Test Image for Checking Wide Gamut Monitors
« Reply #4 on: August 31, 2016, 01:16:29 am »

On my Wide gamut with factory cal the upper set of images are all mure saturated on the left side. What does that mean. On the first test image in this thread the the monitor passes the test nicely.
Logged

Simon Garrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 742
Re: Test Image for Checking Wide Gamut Monitors
« Reply #5 on: August 31, 2016, 04:33:00 am »

On my Wide gamut with factory cal the upper set of images are all mure saturated on the left side. What does that mean. On the first test image in this thread the the monitor passes the test nicely.

Yes I get the same on "wide gamut monitor2.tif".  I'm using two different wide-gamut, home calibrated/profiled monitors - both of which have a colour space wider than Adobe RGB. 

I don't quite understand the significance of having alternating pixels of different colours.  Why not just use solid colours, as here, for example?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #6 on: August 31, 2016, 01:31:12 pm »

On my Wide gamut with factory cal the upper set of images are all mure saturated on the left side. What does that mean. On the first test image in this thread the the monitor passes the test nicely.

The two images are in different colorspaces. My browser (Edge-Win 10) totally hoses them. They are correct viewed side by side in Photoshop though.  I'll put them both in one image in Adobe and repost.

The images should be viewed in a color managed app like Photoshop, not a browser.

Here's a somewhat fuzzy snapshot of how they appear on a CG318. It's a camera snapshot (output referred) so the colors are not accurate but the relative differences between a wide gamut display and an sRGB ish display are pretty close.

Here's a snapshot of my monitor showing them side by side.

« Last Edit: August 31, 2016, 03:30:47 pm by Doug Gray »
Logged

Simon Garrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 742
Re: Test Image for Checking Wide Gamut Monitors
« Reply #7 on: August 31, 2016, 08:03:44 pm »

The two images are in different colorspaces. My browser (Edge-Win 10) totally hoses them. They are correct viewed side by side in Photoshop though.  I'll put them both in one image in Adobe and repost.

The images should be viewed in a color managed app like Photoshop, not a browser.

Here's a somewhat fuzzy snapshot of how they appear on a CG318. It's a camera snapshot (output referred) so the colors are not accurate but the relative differences between a wide gamut display and an sRGB ish display are pretty close.

Here's a snapshot of my monitor showing them side by side.

That's not what I get.  I opened them in Photoshop, using the embedded profile as Working space.  The second (sRGB) image looks much like your snapshot, but with the first one (Adobe RGB), the left side looks darker and more saturated than the right.

I'm not sure why they should look the same. 

Can you explain to me again the logic behind the panels of alternate pixels of different colours please? 

I'm quite sure my monitor is calibrated/profiled correctly - in fact I've tried this on two different dissimilar monitors, both wide-gamut, but calibrated/profiled with different software. 
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #8 on: August 31, 2016, 11:21:03 pm »

That's not what I get.  I opened them in Photoshop, using the embedded profile as Working space.  The second (sRGB) image looks much like your snapshot, but with the first one (Adobe RGB), the left side looks darker and more saturated than the right.

I'm not sure why they should look the same. 

Can you explain to me again the logic behind the panels of alternate pixels of different colours please? 

I'm quite sure my monitor is calibrated/profiled correctly - in fact I've tried this on two different dissimilar monitors, both wide-gamut, but calibrated/profiled with different software.

Algorithm is simple. Let's take the first patch as an example.
1. Zoom in max on the left side. You will see a repeating pattern of 3 highly saturated cyan pixels and 1 pixel that is slightly orange.
2. Convert the Adobe RGB values for the cyan and orange pixels into X,Y, and Z standard CIE colors. These are linear (gamma 1)
3. Average the XYZ values of the 4 pixels.
4. Measure the RGB values of the solid cyan on the right side of the patch. These are all the same.
5. Convert to X, Y, Z.
6. They will match within small rounding errors.

You can use brucelindbloom's excellent site http://www.brucelindbloom.com/ which has a color calculator making it easy to get X,Y,Z values from a color in any standard colorspace.

As for what you are seeing, I would guess that it may be due to monitor interpolation. The monitor must be run at native pixel resolution. Many hi-rez monitors are run at a reduced, simulated resolution to make lettering look larger and they don't properly interpolate color in those modes.


In fact I just checked my CG318 and CG301 monitors. When the monitor is not set to native resolution the colors on the left are significantly more saturated. One way to mitigate this and tell if your monitor (or possibly the OS) is interpolating pixels is to zoom up 800% and let the patch take up most of the monitor. Then back way off of the monitor 10 or 20 feet and you should see the right and left sides much closer to the same color.


The 3 cyan pixels on the left side are highly saturated at the Adobe gamut's edge. When combined with the slightly orange pixel the color saturation is reduced to the edge of the sRGB gamut. If the display is capable of Adobe RGB at that cyan color it will, when combined with the orange pixel, produce the same cyan color as the solid pixels on the right side of the patch which are on the boundary of sRGB.

I decided to check the left/right side colors with Patchtool and they match to about .5 dE1976.

I can think of one other possible reason. It's possible the monitor itself is doing some sort of spacial interpolation to compensate for the fact that the red, green, and blue pixels are slightly offset from each other. There were techniques that did that with text to make fonts look better. There were "fine tuning" dialogs where you were asked to choose the image with the best text clarity. It's entirely possible some monitors do that internally since they know the positions of the RGB elements.  If you have a macro lens and take a closeup of the monitor's pixels at the patch transition it might provide insight. And, if this is the case I can think of a workaround.
« Last Edit: September 01, 2016, 12:40:59 am by Doug Gray »
Logged

Lundberg02

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 379
Re: Test Image for Checking Wide Gamut Monitors
« Reply #9 on: September 01, 2016, 01:07:15 am »

Monitors render perceptual.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #10 on: September 01, 2016, 01:43:45 am »

Monitors render perceptual.
Monitors render in perceptual increments. That's the purpose of gamma > 1.  However, they are profiled as colorimetric devices and colors are rendered colorimetrically in almost all cases.
Logged

Simon Garrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 742
Re: Test Image for Checking Wide Gamut Monitors
« Reply #11 on: September 01, 2016, 04:20:32 am »

My monitors are both running in native resolution.  However, you are right that it is a scaling issue - I'd missed your note to view at 100%, sorry about that. 

An interesting exercise.  What do you think it shows that solid patches of colour would not show?  That's not a critical question: I've used solid colours before myself, and wanted to see how your tests might be add to that.
 
« Last Edit: September 01, 2016, 04:36:06 am by Simon Garrett »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #12 on: September 01, 2016, 09:26:05 am »

My monitors are both running in native resolution.  However, you are right that it is a scaling issue - I'd missed your note to view at 100%, sorry about that. 

An interesting exercise.  What do you think it shows that solid patches of colour would not show?  That's not a critical question: I've used solid colours before myself, and wanted to see how your tests might be add to that.

It's difficult to tell how close a wide gamut monitor comes to achieving full Adobe RGB (1998) using only solid patches without instrumentation to measure it. It could be of use when someone doesn't have a spectro or colorimeter to profile, or even check, a monitor. Or, just to check their color management workflow. It would, for instance, be nice if browsers, which scale images over variable ranges, would properly both color manage and interpolate properly when scaling.

It's also a check on how good image resampling is at retaining proper color. It also might be of some use to check printer drivers which resample to native printer rez.
Logged

Simon Garrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 742
Re: Test Image for Checking Wide Gamut Monitors
« Reply #13 on: September 01, 2016, 09:45:22 am »

Good points. 

I know where the RGB primaries are for my monitors, but as you say some people may not know.  I agree with your points about interpolation, also about browser and other program colour (mis) management.

It's galling that Microsoft go to all the trouble of using colour management in IE and Edge, but don't bother to look up the monitor profile and use sRGB instead.  How many lines of code would it take to make that a user option?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #14 on: September 01, 2016, 10:59:07 am »

Good points. 

I know where the RGB primaries are for my monitors, but as you say some people may not know.  I agree with your points about interpolation, also about browser and other program colour (mis) management.

It's galling that Microsoft go to all the trouble of using colour management in IE and Edge, but don't bother to look up the monitor profile and use sRGB instead.  How many lines of code would it take to make that a user option?

It is galling. I'm not one to tilt with windmills but I do want to know where they are and avoid the blades. 
Logged

Lundberg02

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 379
Re: Test Image for Checking Wide Gamut Monitors
« Reply #15 on: September 04, 2016, 05:01:54 am »

Doug , my monitor renders perceptual it has a perceptual rendering intent, it is wide gamut. It is my understanding that all consumer monitors render perceptual. Why would they clip?  Cite a source.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #16 on: September 04, 2016, 04:00:32 pm »

Doug , my monitor renders perceptual it has a perceptual rendering intent, it is wide gamut. It is my understanding that all consumer monitors render perceptual. Why would they clip?  Cite a source.

Virtually all displays are rendered colorimetrically to the degree feasible and have identical tranforms for Perceptual and Relative Intent. That is, they do not render Perceptual Intent differently from Relative Intent. Still, they are not perfect and they can be considered to render Perceptually near the black point. The biggest limitation with rendering accurately is the monitor's dynamic range. Worse still, is glare from ambient light makes the black point uncertain and, unless the working ambient is quite dark, well above the monitor's native black point. This area is largely ignored in V2 profiles but V4 profiles provide mechanisms to account for black point dynamic range limits. However, V2 profiles remain the most widely used.

http://www.color.org/icc_white_paper_6_v2_and_v4_display_profile_differences.pdf
Quote
ICC version 2 display profiles typically contain only one rendering intent, and this
rendering intent is typically a mixture of perceptual and colorimetric rendering.
For example, most display profiles assume a display black point luminance of
zero (no light whatsoever), and scale the measured display transfer function
accordingly, but then otherwise encode display colorimetry in the profile.

Monitors clip because they are driven by RGB values and, in 8 bit resolution, clipping occurs when a requested color is transformed to the monitor's colorspace resulting in either negative numbers or numbers exceeding 255 in one or more of the RGB channels. ICC protocols truncate negative values to 0 and those above 255 to 255.

Logged

GWGill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 608
  • Author of ArgyllCMS & ArgyllPRO ColorMeter
    • ArgyllCMS
Re: Test Image for Checking Wide Gamut Monitors
« Reply #17 on: September 04, 2016, 08:00:44 pm »

Quote
ICC version 2 display profiles typically contain only one rendering intent, and this
rendering intent is typically a mixture of perceptual and colorimetric rendering.
For example, most display profiles assume a display black point luminance of
zero (no light whatsoever), and scale the measured display transfer function
accordingly, but then otherwise encode display colorimetry in the profile.
I don't know where that Idea comes from. ICC spec. says that matrix profiles are colorimetric. There is no such thing as a mixture of perceptual and colorimetric rendering in a matrix profile if a profile conforms to the ICC specifications.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Test Image for Checking Wide Gamut Monitors
« Reply #18 on: September 04, 2016, 08:15:45 pm »

I don't know where that Idea comes from. ICC spec. says that matrix profiles are colorimetric. There is no such thing as a mixture of perceptual and colorimetric rendering in a matrix profile if a profile conforms to the ICC specifications.

It's a quote from the ICC white paper I linked to.  It just observes that display profiles frequently handle black point issues by scaling colorimetric data from the black point. That makes it not colorimetric in a strict sense. Essentially it is the same as BPC in Photoshop's Rel. Col. These days most monitors have over a 1000:1 dynamic range so the effect is tiny at most. At least if one doesn't create monitor profiles using higher BPs but that has to be done intentionally.
Logged
Pages: [1]   Go Up