LUT has "broken" blue

Home Forums Help and Support LUT has "broken" blue

Viewing 14 posts - 1 through 14 (of 14 total)
  • Author
    Posts
  • #2799

    Rodrigo Silvestri
    Participant
    • Offline

    Hi. I’m a colorist (film and video postproduction) and I’ve been using dispcalGUI since 2010 for profiling and calibrating video monitors and some computer displays. I have two old 1iDisplay2s and one i1DisplayPro.

    I normally use DisplayCAL to the “Video 3D LUT for Resolve” setting, create a ~700 patches testchart, then profile the monitor, and finally I create a LUT with a custom made ICC profile using the Synthetic ICC tool and the 3DLUT Maker. Then I check the LUT in Resolve to see if it’s visually alright, and then perform some tests with HCFR.

    This week, while trying to update the calibration LUTs for two video monitors I encountered a weird issue, and cannot find where it comes from. When checking a color ramp chart after applying the LUT, I see that the blue saturation scale is weird. I uploaded the original chart and a picture of the problem (taken with my phone, totally off any kind of similarity, but you can see what’s happening).

    The rest of the LUT is perfect, greyscale is calibrated, gamma is OK, red, green and blue patches are in their respective places (one of the monitors has a wide gamut and I’m reducing it to Rec709).

    I tried reducing my procedure (using standard Rec709 instead of synthetic ICC, use auto-optimized testcharts, use different testcharts (~200 to ~1500 patches), resetting the monitors settings and double-checking that there are no LUTs loaded anywhere.

    I did this procedure in 2 different OSX computers, both running DisplayCAL 3.1.3.1, I tried the old and the new probes. The monitors I’m using are connected via SDI and HDMI through Blackmagic video cards, and have no internal LUTs, no dynamic contrast settings or anything similar. I have created many profiles for these monitors with no issues in the past, with the same color meters.

    It happened once or twice that I could create an OK LUT using small testcharts (~300 patches) and using the “Create 3D LUT after profiling” option in DisplayCAL, but then I would use the 3DLUT Maker with the same settings, or with my custom ICC profile, and both would have this issue in the blue hue.

    I’d appreciate any help or hints of what could be happening here.

    Thanks,

    Rodrigo Silvestri

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

    Calibrite Display Pro HL on Amazon  
    Disclosure: As an Amazon Associate I earn from qualifying purchases.

    #2803

    Florian Höch
    Administrator
    • Offline

    Hi,

    please use the “Create compressed archive…” button (next to “Settings”) with the profile selected and send me/attach the resulting file. Thanks.

    #2809

    Rodrigo Silvestri
    Participant
    • Offline

    Here! 🙂

    Attachments:
    You must be logged in to view attached files.
    #2811

    Florian Höch
    Administrator
    • Offline

    Thanks. The effect you’re seeing is because a portion of the Rec. 709 blue region is outside of the specific display’s gamut, and so it clips.

    #2812

    Rodrigo Silvestri
    Participant
    • Offline

    Thanks for the reply.
    I would expect the software to handle this clipping differently. It appears as a very clear patch in the display when applying the LUT (as you saw in the picture). Actually, if the display has some kind of bad color matrix that causes the patch, I would expect the LUT to reduce it, but not to generate it from nothing.

    This never happened to me before, even with previous versions of dispcalgui, and is happening now on two different displays. I insist on asking if there could be anything else to dig.

    Thanks,

    Rodrigo.

    #2813

    Florian Höch
    Administrator
    • Offline

    It’s a display hardware limitation, there’s nothing that can be done about it. Your display’s gamut is simply not large enough to encompass all of Rec. 709.

    #2814

    Rodrigo Silvestri
    Participant
    • Offline

    I’m not using rec709 to create my LUTs. In fact, I just created a synthetic profile with the primaries measured from the display and it does the same thing. Do you see the weird band that I’m talking about in the picture? That’s not normal for dispcalgui.

    Edit: I mean, I know I can use a monitor with a small colorspace and make a LUT using absolute colorimetric rendering intent to match, and it will leave a big part of the gamut clipped, but the colors inside the display’s gamut should be represented just fine.  I’ve done it in the past (with testing purposes only). This display has its blue point slightly different than REC709, so yes, it should clip out a part, but it’s doing a big patch of desaturated/saturated blue where it shouldn’t.

    I just edited the same picture to (somehow) match what I see visually in the display after applying the LUT. The blue saturation scale is broken. Can anyone explain what’s happening here?

    Thanks,
    Rodrigo.

    Attachments:
    You must be logged in to view attached files.
    #2815

    Florian Höch
    Administrator
    • Offline

    This is exactly what I would expect given that your display does not behave in a perfectly additive way, i.e. the synthetic profile that you created (which in contrast is perfectly additive) should have a gamut that is larger than the actual display gamut. See the attached gamut comparisons. In both, the colored shape is your actual display gamut, and the wireframe in one case is Rec. 709 and in the other a synthetic profile created from just the display primaries and white. In both cases, a significant portion of the source gamut (wireframe) lies outside your display gamut (clipping).

    Attachments:
    You must be logged in to view attached files.
    #2820

    Rodrigo Silvestri
    Participant
    • Offline

    This LUT is desaturating a portion of the gamut that is inside the display’s gamut with no purpose, and making it look really bad. As I said before, I’ve been doing this with dispcalgui and other softwares for a long time and this is the first time that I see something this bad. Why didn’t this happen before?

    My main intention with this monitor is to make a LUT to balance its grayscale and limit it to rec709 and another to imitate P3 (I know it will not be close enough but I want colors inside the displays gamut to be represented as good as possible).

    Is there any way to make a LUT that doesn’t destroy the blue saturation or any other part of the gamut, that just clips out the portion that is outside the gamut without making it look bad?
    Do you recommend me another way of doing this with dispcalGUI?

    Thanks,

    Rodrigo

    #2821

    Florian Höch
    Administrator
    • Offline

    This LUT is desaturating a portion of the gamut that is inside the display’s gamut with no purpose

    The data you provided doesn’t indicate this at all, so I’m a bit puzzled how you come to this conclusion.

    and making it look really bad

    The aim of accurate calibration is never to make something subjectively “look good”. The aim is to reproduce colors as specified and as accurately as possible, and if the display can’t do that due to its limitations, then the resulting LUT will show this.

    Is there any way to make a LUT that doesn’t destroy the blue saturation or any other part of the gamut

    The LUT doesn’t “destroy” anything, it fixes all in-gamut colors as expected. Anything that lies outside is naturally clipped. If you don’t want clipping,  you’ll have to get a better display with a larger gamut, so that the desired source gamut fits inside.

    Do you recommend me another way of doing this with dispcalGUI?

    There is no software in the world that will make a display magically achieve a larger gamut than it actually has.

    #2822

    Rodrigo Silvestri
    Participant
    • Offline

    There is no software in the world that will make a display magically achieve a larger gamut than it actually has.

    Yes, I know this, I’d be thankful if you could reread what I wrote, because I didn’t intend a LUT to do that and I think you’re misunderstanding me.

    1. Set aside the rec709 blue point and any kind of standards, just with the purpose of understanding how to use youe software, is there a way to make a LUT that limits  my display to the rec709 green and red primaries (and balances its grayscale) with a different blue point so that it doesn’t create a patch in the saturation scale?
    2. My monitor’s green point is “bigger” than Rec709 but doesn’t reach the P3 green. The difference between the green is huge compared to the blue difference, which is close enough for my purposes. If I use this profile to make a LUT to imitate P3, it still creates this same patch in the blue saturation scale but the green looks perfect, there’s no step in its saturation scale. It does with the green what I would expect with the blue too. Why does this happen?

    Thanks again,
    R.

    #2828

    Florian Höch
    Administrator
    • Offline

    is there a way to make a LUT that limits my display to the rec709 green and red primaries (and balances its grayscale) with a different blue point so that it doesn’t create a patch in the saturation scale?

    The main problem is not necessarily the blue chromaticity, but that your display’s blue is too dark. To fit any matrix-based space (which both Rec. 709 and P3 are) completely inside your display gamut, you’d have to make the former considerably smaller, and you can’t just change one of the matrix’ primaries because they need to add up to white otherwise you’ll introduce a color cast in neutrals.

    It does with the green what I would expect with the blue too. Why does this happen?

    There’s nothing “special” happening with the green. On one hand the human visual system is much less likely to notice errors in very bright saturated colors like green (which is the brightest of the primaries). But most significantly, in case of your specific display, the green difference towards P3 is mainly one of hue and chroma, not lightness, so it clips to the closest point with about the same lightness, while for a large part of the blue region the errors are in chroma and lightness, clipping towards the darker blue of your display.

    What you can try to combat the effect is to reduce the LUT size thus increasing interpolation (but decreasing accuracy).

    I’d also like to point out that visual errors due to display limitations that are often only apparent in such specially crafted test images are usually not reflected as much (or at all) in real-world imagery.

    #2847

    Rodrigo Silvestri
    Participant
    • Offline

    Ok, thanks.

    I remember that with the first versions of dispcalgui that supported exporting 3D LUTs, I did make a LUT that would only balance the grayscale of the monitor without touching saturation or color space. I think (I can’t recall exactly) I did that by selecting the same source and target profile.

    Now, if I try to do that, the 3DLUT Maker shows an error saying that the source and target profile are the same (indeed!). How can I, now, create a LUT that does only that, balance the grayscale without touching anything else?

    Thanks,
    R.

    #2848

    Florian Höch
    Administrator
    • Offline

    I think (I can’t recall exactly) I did that by selecting the same source and target profile.

    That would create a “null” transform, i.e. one that doesn’t change the input values at all (that’s why newer versions check if source and destination are the same).

    How can I, now, create a LUT that does only that, balance the grayscale without touching anything else?

    It should be relatively straightforward to convert an Argyll CMS .cal file to .cube (i.e. just copying the values between BEGIN_DATA..END_DATA in the .cal file, then deleting the “index” column should work), but you’d have to do it manually.

Viewing 14 posts - 1 through 14 (of 14 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS