Welcome to PhantomPilots.com

Sign up for a weekly email of the latest drone news & information

Feedback on this 3D LUT

Discussion in 'Pro/Adv Discussion' started by MikeyOnline, May 6, 2016.

  1. MikeyOnline

    Joined:
    Apr 22, 2016
    Messages:
    338
    Likes Received:
    116
    Been fooling with creating 3D LUT's for the P3P in D-Log mode. I've attached a .cube file if anyone wants to take a look to see how it works for them. This thing started life as an ICC profile made from a color target shot with the P3P camera, so I was going for "neutral" colors. Seems to work pretty well (I've been applying it in ColorDirector but it also works in Davinci Resolve). Just curious what you guys think: those of you who are familiar with applying 3D LUTs. This is primarily just to see if I'm on the right track and get some feedback.

    Thanks,
    Mike
     

    Attached Files:

  2. MikeyOnline

    Joined:
    Apr 22, 2016
    Messages:
    338
    Likes Received:
    116
    OK, here's another one that's a bit better.

    Now, what I've noticed is that as far as I can tell, it's not possible to create a color grading (3D LUT) that is "accurate" for D-Log under various exposures. Both of these 3D LUTs started life as a color profile created from a ColorChecker SG (first version) and an IT8 target (second version). Those profiles were then converted to 3D LUT cube transforms so in theory, they should produce reasonably accurate color and tone curves.

    Here's the problem I've found. Over many years of developing ICC profiles (and profiling software) for cameras, I know that it isn't possible to develop a good ICC profile for a device unless you can apply that profile (or LUT) to the raw colors. And D-Log is not raw! It has already been processed. Trying to apply a profile or LUT to a processed image or video basically means that the designed LUT will only apply (properly) to the condition of the video where the color target was shot because I'm finding that the tone curves change depending on the exposure of the camera. And I don't mean just darker/brighter, I mean the actual shape of the curve. To be able to develop a consistent LUT, we'd need to be able to access the raw (linear) RGB values from the video which, as far as I can tell, is not possible. As it stands, I'm finding that if you have one 3D LUT that seems to work well and then you (or the AC) changes exposure based on something else in the frame, that LUT will not work well. In addition, it appears that all color values in each video are 8 bits (0-255) so storing the video in a wider gamut like D-Log has its down sides as well: bringing that back to good color can sometimes produce banding in soft gradients like sky.

    Given what I've found, I'm just switching mine back to "None" for color mode as that seems to render the mose consistent color. I can see LUTs being useful in a professional environment where you have a boatload of LUTs designed for different shooting conditions and you want to take the time to learn which LUT is applicable to which lighting conditions and exposure, but for me (the hobby shoots), it's just too much trouble.

    Anyway, thought I'd share.

    Mike
     

    Attached Files: