09:58 | polyrhythm | We have an email thread going but here is the latest on color calibration...:
| |
09:58 | polyrhythm | we're close.
| |
09:58 | polyrhythm | https://i.imgur.com/6UCpRhY.png
| |
09:58 | polyrhythm | A pretty crude color match from Resolve indicates we're in the ballpark, even with the bare minimum 6500K calibration I did just now
| |
09:58 | polyrhythm | The cyan is pretty far off at 13% but everything else seems pretty acceptable
| |
10:00 | polyrhythm | this is really not going to be perfect anyway with just a 3x3 matrix, so thos numbers are pretty good.
| |
10:00 | polyrhythm | the next steps would be solving some of the issues with our current photos, getting spec-compliant raw2dng, trying out the IT8 chart to see how it fares with calibration, etc
| |
10:00 | polyrhythm | overall I'm pleased with the initial attempt
| |
10:01 | polyrhythm | most crucially, it looks like we achieved a good neutral calibration with no real twisted hues or garbage like that, just some saturation/luminance diffs. very easy to work with
| |
10:51 | se6astian | great, thanks for the updated
| |
10:55 | polyrhythm | oh and I just noticed I used the "Legacy" option for the color chart in Resolve, whereas the correct version is the non-legacy option, so maybe I'll try again later and see if it improves the cyan matching
| |
12:32 | vup | polyrhythm: what does the percentage in the resolve color match thingy mean?
| |
12:50 | BAndiT1983 | maybe deviation
| |
12:53 | vup | yeah, but under what measure?
| |
13:09 | BAndiT1983 | as i've missed majority of discussions, am not that deep into what was targetted, what would be the measure in this case? am seeing the x-rite classic (legacy) color chart and the deviation after correction
| |
13:10 | BAndiT1983 | or ist it about the auto values there?
| |
13:16 | vup | BAndiT1983: no I think you have all the details. But with measure I mean you have a triplet of correct (X, Y, Z) values for the color checkers patches and a triplet of the (X, Y, Z) values the camera gives. How do you now combine / compare these two triplets to get a single percentage difference?
| |
13:21 | BAndiT1983 | from my very basic knowledge it's just known values for the color chart and then the matrix for the conversion, the values are compared after applying the matrix, that would be it, in my understanding, this is where i miss a chunk of infos probably and have to look through the meeting notes, when my personal things are sorted out after this week
| |
13:23 | BAndiT1983 | am refering to something like this -> https://www.imatest.com/docs/colormatrix/ and from trials with display calibration
| |
13:38 | intrac | left the channel | |
13:38 | intrac | joined the channel | |
14:34 | vup | BAndiT1983: yes of course, but how "exactly" are they compared
| |
14:34 | vup | hmm the light used seems to have different angular dependencies for different frequency components: https://f.coroot.de/cmv12k_color_response/greycard_zoomed.pdf
| |
14:35 | BAndiT1983 | interesting
| |
14:42 | aombk2 | joined the channel | |
14:43 | aombk | left the channel | |
17:00 | se6astian | MEETING TIME, who is here?
| |
17:02 | se6astian | BAndiT1983 messaged me is is currently away but he reported he is still working on panel creation outline tuning, here is a work in progress screenshot
| |
17:02 | se6astian | https://i2.paste.pics/9084733a3ac09256e71ee6b1f8effc8f.png
| |
17:05 | se6astian | there is quite a lot going on in the color characterization front, but maybe someone wants to talk about it who is directly involved?
| |
17:08 | se6astian | anyone else around who wants to report/share?
| |
17:11 | se6astian | ok then, MEETING CONCLUDED
| |
17:44 | McUles[m] | left the channel | |
17:44 | davidak[m] | left the channel | |
17:44 | underpantsgnome[ | left the channel | |
17:49 | McUles[m] | joined the channel | |
18:02 | underpantsgnome[ | joined the channel | |
18:02 | davidak[m] | joined the channel | |
21:19 | jn | left the channel | |
21:19 | jn | joined the channel | |
21:20 | jn | left the channel | |
21:20 | jn | joined the channel | |
23:10 | anuejn | this looks like some pretty nice hardware as an axiom-remote interim: https://carthing.spotify.com
| |
23:10 | anuejn | also it is easy to run custom code (web stuff) on it: https://carthing.spotify.com
| |
23:11 | anuejn | which would mean that it cauld be a low-effort solution for controling the axiom in the short term :)
| |
23:12 | anuejn | though sadly it is not very available in europe :(
| |
00:17 | polyrhythm | vup: The percentage means the DeltaE between the final matched color and the reference value of the patch. In this case, my test was to take the calibrated camera into Resolve and see how easily it could match the image, with the logic that a calibrated image is closer to baseline than an uncalibrated one and thus easier to match. It's a little misleading because I'm implying that my footage directly is those values; it's just the result after I let
| |
00:17 | polyrhythm | Resolve have a go at the calibrated image. The reason is that imaging pipelines throw the DNG into different provessing color spaces, so obtaining some kind of absolute DeltaE between reference patch values and the color chart photo is not super useful unless you know every step the processing software is doing when it renders the image from a DNG -> sRGB display
| |
00:19 | polyrhythm | in other words, I don't think there's value in taking a calibrated DNG, putting it into Nuke or Resolve or whatever, and just checking the DeltaE directly with reference values, because those software are going to process the DNG differently and put them into different working spaces to render them. I guess Nuke is the most explicit in this case in what is happening
| |
00:20 | polyrhythm | as best as I can tell, the calibration can be considered successful is we get ballpark correct images from the DNGs using minimal processing to a display transform and without hue twisting or significant saturation and luminance deltas
|