Colorimetrically accurate f.lux/Redshift?

Home Forums General Discussion Colorimetrically accurate f.lux/Redshift?

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #1747

    anonymous SourceForge
    Member
    • Offline

    I was wondering if it’d be easy to add white balance shifting over time to dispcalGUI, similar to f.lux, Redshift and other tools, but to do so in a colorimetrically accurate way (shift the white point in XYZ space), with the option to shift towards an absolute kelvin value (sun or black body) or to shift relative to device white.

    #1748

    Florian Höch
    Administrator
    • Offline

    That would require re-measuring or at the very least re-calculating the profile after each whitepoint shift, which is a lengthy process, so I don’t think it’s feasible.

    #1749

    anonymous SourceForge
    Member
    • Offline

    A poor man’s solution could be to have a “day profile” and a “night profile” and to just do a linear interpolation between the two LUTs for the intermediate steps?

    #1750

    Florian Höch
    Administrator
    • Offline

    I’ve looked into it, but it’s more or less a gimmick – adjusting whitepoint via calibration curves leads to reduced gamut when the channels are reduced, and I wouldn’t recommend using a display that doesn’t have a “daylight”-like whitepoint for anything remotely color-critical, it’s just not going to work. It’s possible to get fairly accurate whitepoint and graybalance adjustment though.

    I’ve also looked at f.lux, and it has a few shortcomings:

    • It uses the active calibration curves (good), but doesn’t use the display profile colorimetry (I assume internally it uses sRGB) so the whitepoint mapping is inaccurate like you said. For the purpose it’s meant to be used, that’s probably ok though.
    • Calibration curves are always quantized to 8 bit by f.lux, which leads to banding.
    #1751

    Florian Höch
    Administrator
    • Offline

    As an update, I’ve added the functionality to alter an existing profile/calibration whitepoint in the latest beta snapshots (3.0.0.x), but it doesn’t have any UI (and I’m probably not going to add one). It’s an computationally expensive operation though, so there won’t be real-time shifting.

    Usage is as follows:

    ~~~~~~
    0install run –not-before=3.0.0.4
    –command=change-display-profile-cal-whitepoint —
    http://dispcalgui.hoech.net/0install/dispcalGUI.xml
    [-t temp | -T temp | -w x,y] [–cal-only] [inprofile]
    outfilename
    ~~~~~~

    Example 4000K daylight from the current display profile (needs to be ICCv2):

    ~~~~~~
    0install run –not-before=3.0.0.4
    –command=change-display-profile-cal-whitepoint —
    http://dispcalgui.hoech.net/0install/dispcalGUI.xml
    -t4000 outfilename
    ~~~~~~

    • This reply was modified on 2015-05-27 22:02:30 by fhoech.
Viewing 5 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS