Measuring Color Accuracy on Smartphones

Home Forums General Discussion Measuring Color Accuracy on Smartphones

Viewing 13 posts - 16 through 28 (of 28 total)
  • Author
    Posts
  • #30479

    Vincent
    Participant
    • Offline

    RGB numbers are the same and rendered as raw RGB numbers into the screen*, then measured against “reference” whetever it is.

    *) an exception is verification “validate if profile is accurate enough to render accurately colors of some particular image colorspace in Ps/Lr (simulation profile)”

    Again, no HTML report no support since we cannot know what you did wrong.

    • This reply was modified 2 weeks, 5 days ago by Vincent.
    #30482

    Vincent
    Participant
    • Offline

    *) an exception is verification “validate if profile is accurate enough to render accurately colors of some particular image colorspace in Ps/Lr (simulation profile)”

    And that require a display profile to transform source colorspace (simulation) to display colorspace

    #30483

    Lenk83
    Participant
    • Offline

    Got you. Please see attached the report for the iphone 12 pro. Results are pretty bad with display p3 measure profile. But with sRGB they are good (around 0,74 delta E).

    • This reply was modified 2 weeks, 5 days ago by Lenk83.
    Attachments:
    You must be logged in to view attached files.
    #30486

    Vincent
    Participant
    • Offline

    Lower 2Dplot, CIE a*b*. Browser seems to be running color management transforming untagged HTML colors to sRGB.

    Since grey is on spot, try to generate RGBCMY patches in GIMP or PS, 255 value in a DisplayP3 image (assign profile or create image in that colorspace). Then save tiff or png embeding image profile, so phone’ image color viewer “can”, if it has some basic color management that honors content, send unaltered P3 color paches to screen.

    Measure those patches with HCFR free measurement and set Display P3 as reference colorspace.
    You can also use spotread on commandline and compare deltaE to DisplayP3 references by yourself, instead of HCFR, but better with automation.

    • This reply was modified 2 weeks, 5 days ago by Vincent.
    • This reply was modified 2 weeks, 5 days ago by Vincent.
    #30490

    Lenk83
    Participant
    • Offline

    Since grey is on spot, try to generate RGBCMY patches in GIMP or PS, 255 value in a DisplayP3 image (assign profile or create image in that colorspace). Then save tiff or png embeding image profile, so phone’ image color viewer “can”, if it has some basic color management that honors content, send unaltered P3 color paches to screen.

    Well, that seems to help. On HCFR I got with the iPhone 12 Pro Delta E of 0,78 with dci p3 and 0,79 with sRGB, for each I generated a set of color patterns sRGB and P3 as well.

    The Samsung Galaxy S21 Ultra gets 1,22 on P3 and 1,11 on sRGB.

    The much cheaper and technically by far not good ZTE Axon 20 gets 3,46 and 3,42.

    The values seem very plausible on first sight for me.

    Thank you a lot for that advise. Still, im not sure if that is a good way to go since I am only testing six colors this way and not about 51 or more like DisplayCAL does. Are six enough to make an educated assumption about display color accuracy?

    #30510

    Lenk83
    Participant
    • Offline

    Now I have a problem with a device (iPad Pro) which has a high delta E on white and a temperature of 6700k, little too much blue delta E too. All on dci p3 on hcfr (created with photoshop, tried with different apps from photos to photoshop and Lightroom – I wish so hard there would be a displaycal mobile app with cms)

    As far as i understand, temperature has nothing to do with the delta E itself. The white is still a clear white no red or yellow tones, you could describe it as a cool white. But that does not explain the high delta e (around 3, while other colors are around 0,5-1).

    Furthermore, if I get things right, the measured white value directly influences the results of the other colors.

    Since the new iPad uses mini LED display technology and is know for high accuracy especially in P3, I assumed this  measurement is corrupt because of wrong correction. I tried different corrections in LED but only got slightly different results.

    In the end i measured the exact CIE values of the displays white and put these as targeted whitepoint instead of D65. That gave way better results. Is this generally the way to go or do I miss something? It is unsatisfying to me because in this way I would not honor an accurate white value and furthermore  ignore if a display shows a reddish or yellowish or blueish white and lacks of precise white

    again: would so love to see a displaycal mobile app since i like that Programm much more that hcfr.

    • This reply was modified 2 weeks, 2 days ago by Lenk83.
    • This reply was modified 2 weeks, 2 days ago by Lenk83.
    • This reply was modified 2 weeks, 2 days ago by Lenk83.
    #30514

    Vincent
    Participant
    • Offline

    Now I have a problem with a device (iPad Pro) which has a high delta E on white and a temperature of 6700k, little too much blue delta E too. All on dci p3 on hcfr (created with photoshop, tried with different apps from photos to photoshop and Lightroom – I wish so hard there would be a displaycal mobile app with cms)

    As far as i understand, temperature has nothing to do with the delta E itself. The white is still a clear white no red or yellow tones, you could describe it as a cool white. But that does not explain the high delta e (around 3, while other colors are around 0,5-1).

    IDNK what you are talking about. Assumed white point + blue primary and all others in place?

    Furthermore, if I get things right, the measured white value directly influences the results of the other colors.

    Measured white is the a*b* neutral axis reference for greys… as it should be. If you want to compare against actual values move combo to xyY and abs values in DisplayCAL HTML report without PCS translation.

    Since the new iPad uses mini LED display technology and is know for high accuracy especially in P3, I assumed this  measurement is corrupt because of wrong correction. I tried different corrections in LED but only got slightly different results.

    No Apple screen is known for that, specially on whitepoint. It’s typical to be moved towards b- (if you use D65 as reference b=0), bluish non D65 white and within 2-4dE radius in “color” plane from daylight reference. Several samples in user reports in this forum.

    If you have an Xrite spectrophotometer or can use one, make sure to use 3nm mode and plot spectral power distribution. If it uses some PFS phosphor variant (double spike in red) device may be not accurate enough to measure it even at 3nm so 2-3dE to white target reference white using 3nm mode seems to me “ON SPOT” (OK) measuring a PFS.

    In the end i measured the exact CIE values of the displays white and put these as targeted whitepoint instead of D65. That gave way better results. Is this generally the way to go or do I miss something? It is unsatisfying to me because in this way I would not honor an accurate white value and furthermore  ignore if a display shows a reddish or yellowish or blueish white and lacks of precise white

    I understand now, it’s a HCFR report. Ref values are absolute instead of calculate grey “color” (a*b*) against measured white.

    You should run all test using content reference colorspace, for example a monitor against D65 and 2.2g or sRGB trc.
    If you wish to provide an additional report to your readers about grey neutrality, like prad.de in deltaC+ grey range plot, you can do what you explain… but it is an additional test. It is not a test validating if display matches some reference colorspace. IDNK if HCFR user interface plots values in a*b* but this kind of test of grey neutrality it’s easier to read and understand that way, measured white to be used as reference with a*=b*=0.

    again: would so love to see a displaycal mobile app since i like that Programm much more that hcfr.

    There is a paid app, an ArgyllCMS mobile app. It’s expensive, use the demo app to make sure device is supported and such.
    It’s a full app: measures, computes distance… etc. It is not a web hook like Calman or Lightspace that is listening on a computer to devices wanting to connect.

    If you wish an app like calman lightspace client ask in AVS Forum, HCFR thread. Dev is Dominic Chan (I do not remember). Maybe HCFR already has some equivalent feature like that calman mobile app, IDNK. Maybe HCFR devs are willing to add it.

    • This reply was modified 2 weeks, 2 days ago by Vincent.
    #30517

    Lenk83
    Participant
    • Offline

    No Apple screen is known for that, specially on whitepoint. It’s typical to be moved towards b- (if you use D65 as reference b=0), bluish non D65 white and within 2-4dE radius in “color” plane from daylight reference.

    Exactly what I have here. But it is not about Greyscale. Its about the primary and secondary colors (RGB, CMY). If you do a measurement for these, the program will also measure the white pattern in the end. After having the white pattern measured, all the other measurements change according to that white point. And that’s what’s confusing me. If it would have a better whitepoint, all the deltaE get better. I tried it by replacing the D65 target with the real whitepoint of that display. So color accuracy with hcfr, it seems, depends on white point.

    #30518

    Lenk83
    Participant
    • Offline

    May it be, that Apple is pushing the whitepoint beyond 6500k in order to achieve a grey linearity? So they sacrifice a bit of color accuracy in order to get a perfect greyscale?

    #30519

    Lenk83
    Participant
    • Offline

    And its getting more confusing. Putting the whitepoint issue aside.

    1. Problem:DisplayCAL gives better results with sRGB and Greyscale than HCFR
      Possible explanation: Maybe because of the missing or wrong CMS in safari browser which I use on display cal, not on hcfr.
    2. Problem: iPhone 12 Pro seems to have the more accurate display according to hcfr, which is odd because the iPad is promoted as high accurate for graphical purposes (I know, they promise a lot in commercials what is not true.
      Possible explanation: No clue. Maybe the iPhone 12 Pro DOES have the better screen for color accuracy.

    See the attached overview. All measurements from iPad Pro, only to the right the iPhone. Whats your opinion about it?

    Attachments:
    You must be logged in to view attached files.
    #30521

    Vincent
    Participant
    • Offline

    May it be, that Apple is pushing the whitepoint beyond 6500k in order to achieve a grey linearity? So they sacrifice a bit of color accuracy in order to get a perfect greyscale?

    No, just cost less to them. Image quality is not part of their goals.

    #30522

    Vincent
    Participant
    • Offline

    And its getting more confusing. Putting the whitepoint issue aside.

    1. Problem:DisplayCAL gives better results with sRGB and Greyscale than HCFR
      Possible explanation: Maybe because of the missing or wrong CMS in safari browser which I use on display cal, not on hcfr.
    2. Problem: iPhone 12 Pro seems to have the more accurate display according to hcfr, which is odd because the iPad is promoted as high accurate for graphical purposes (I know, they promise a lot in commercials what is not true.
      Possible explanation: No clue. Maybe the iPhone 12 Pro DOES have the better screen for color accuracy.

    See the attached overview. All measurements from iPad Pro, only to the right the iPhone. Whats your opinion about it?

    Plot deltaC across greyscale, like in a prad.de review. Watch out for grey range as overall noticiable color cast on grey (one grey to more pink, other to more green, look for a* maxium oscillation). Low range is better than overall lower dC to reference because with very low range grey looks neutral.

    • This reply was modified 2 weeks, 1 day ago by Vincent.
    #30524

    Алексей Коробов
    Participant
    • Offline

    Have you checked browser CMS settings? Type special command in address string (chrome://flags for Chrome, about:config for Firefox, find others in the net) and search for “color” keyword (gfx.color in Firefox). By example, Firefox has choice of color rendering. Here you can also set sRGB assignement to images or all content etc. Find out settings description in the net.

Viewing 13 posts - 16 through 28 (of 28 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS