Current Server Time: 00:01 (Central Europe)

#apertus IRC Channel Logs

2022/09/23

Timezone: UTC


01:30
Spirit532
left the channel
01:30
Spirit532
joined the channel
02:02
polyrhythm
joined the channel
07:25
polyrhythm
left the channel
12:25
polyrhythm
joined the channel
12:29
polyrhythm
hi. i'm ryan, doing some color science. i wanted to share a little progress...
12:29
polyrhythm
1) i figured out that raw2dng's default settings produce incorrect dngs from raw12 because the bayer pattern is wrong.
12:29
polyrhythm
default is 2, you want 1. that sucked up a couple days of frustration.
12:29
polyrhythm
2) the current means of producing the default camera calibration matrix is hacky. we were using a color matching algorithm and just naively matching a test chart shot by a nikon dslr. what we want to do is properly profile it using spectral data of the test chart against a linearized image.
12:29
polyrhythm
i am not done with step 2 yet but i have gotten my head around how raw2dng works in general and have gotten a new matrix into the dng.
12:29
polyrhythm
https://i.imgur.com/xDVO2F4.jpg
12:29
polyrhythm
left is uncalibrated, right calibrated
12:29
polyrhythm
the reds are twisted a bit to orange to me. i suspect this is because I'm assigning the illuminant as "d50" but i in fact have no idea what the illuminant is, so it's probably not really a d50 light.
12:29
polyrhythm
3) right now we are only supplying one camera matrix, under one (unknown) illuminant; this is not correct. we need to supply two camera matrices under two different illuminants (e.g. d50, tungsten) so that processing software can lerp between the two depending on white balance. to-do.
12:30
polyrhythm
4) i sent an email to the team with a larger proposal about color science and what it would take to make this thing ready to actually shoot in the real world with a log curve and a gamut designed to encompass the values coming out of the sensor.
12:31
polyrhythm
oh also, i don't have the damn camera, so it makes all of this a lot harder.:)  i know that it's expensive and not easy to ship off randomly, but it means i have to ask politely if i need images instead of just getting in there and taking photos and iterating, so i am slowed down by that. for now i am making due with the available shots but not
12:31
polyrhythm
knowing the precise nature of the lights is hurting me
12:34
polyrhythm
that is all. i will try to make progress as i can. thanks!
14:50
polyrhythm
left the channel
19:09
se6astian
very interesting ryan, many thanks for the updated and insights
19:09
se6astian
I will also reply to your email now
19:10
se6astian
we discovered a bug in the software that writes the raw images inside the axiom beta some time ago, it was messing up the order in which pixels were written
19:10
se6astian
so we made raw2dng able to develop different bayer pattern orders
19:10
se6astian
I assume you developed an image that still suffered from this wrong order
19:11
littlebobeep
joined the channel
19:11
se6astian
I will try to coordinate collecting the next set of images with better info about the light situation
19:12
littlebobeep
left the channel
23:27
polyrhythm
joined the channel
23:27
polyrhythm
left the channel
23:42
Menkalinan
joined the channel