Current Server Time: 10:23 (Central Europe)

#apertus IRC Channel Logs

2015/02/01

Timezone: UTC


00:07
TD-Linux
left the channel
00:20
TD-Linux
joined the channel
00:30
dmjnova1
joined the channel
00:32
dmjnova
left the channel
00:35
dmjnova1
left the channel
00:37
daFred
left the channel
00:39
intracube_
changed nick to: intracube_afk
01:43
fsteinel
left the channel
01:43
Juicyfruit_
left the channel
01:43
mars_
left the channel
01:44
fsteinel_
joined the channel
01:52
Juicyfruit_
joined the channel
01:52
mars_
joined the channel
02:32
intracube_afk
left the channel
02:50
fsteinel_
changed nick to: fsteinel
04:07
aombk
left the channel
04:21
dmjnova
joined the channel
04:34
davidak
left the channel
04:41
Bertl
off to bed now ... have a good one everyone!
04:41
Bertl
changed nick to: Bertl_zZ
08:24
niemand
joined the channel
08:45
TD--Linux
joined the channel
08:45
TD-Linux
left the channel
08:45
TD--Linux
changed nick to: TD-Linux
08:45
TD-Linux
left the channel
08:45
TD-Linux
joined the channel
09:13
se6astian|away
changed nick to: se6astian
11:14
Bertl_zZ
changed nick to: Bertl
11:14
Bertl
morning folks!
11:20
se6astian
hello!
11:50
aombk
joined the channel
12:01
lab-bot
sebastian created T271: premade Set of LUT/curves. http://lab.apertus.org/T271
13:31
niemand
left the channel
13:49
intracube
joined the channel
14:52
niemand
joined the channel
15:29
Juicyfruit_
changed nick to: Juicyfruit
15:29
Juicyfruit
left the channel
15:29
Juicyfruit
joined the channel
17:38
Bertl
off for now ... bbl
17:38
Bertl
changed nick to: Bertl_oO
20:17
philippej_fr
joined the channel
20:18
philippej_fr
anyone around to discuss what is exactly done to RAW data "as they say" before it gets outputed from traditional camera. Or in other words, what is really raw ?
20:19
philippej_fr
I hope Bertl_oO will be able to help :-)
20:20
philippej_fr
or said otherwise, what is hidden in proprietary cameras?
20:28
philippej_fr
left the channel
20:30
philippej_fr
joined the channel
20:38
danieel
usually the filtering and sensor specific normalization of data
20:39
se6astian
FPN is already corrected in raw formats
20:39
se6astian
dead pixels are either corrected or supplied as meta information
20:40
se6astian
matrix is usually provided as meta information
20:42
philippej_fr
thanks
20:42
philippej_fr
also I was thinking about sensor sepcific features
20:42
se6astian
for longer exposures it might be a complete black frame substraction that is subtracted instead of just a standard FPN profile
20:43
philippej_fr
like analog gain
20:43
danieel
that is LG/HG tonemapping in dual-adc systems
20:43
philippej_fr
or any other trick we don't even know
20:43
philippej_fr
like plr
20:43
philippej_fr
black shading
20:43
se6astian
analog gain happens on the sensor
20:43
se6astian
you only get digital data
20:44
philippej_fr
we are trying to find what others are doing that might impact image and where dop has no control as a sale argument for open source :-)
20:45
se6astian
well the fact alone that we do not know what is happening is enough :)
20:46
se6astian
maybe they do noise reduction
20:46
se6astian
who knows...
20:46
philippej_fr
yes but trying to guess is a goog exercicse
20:46
philippej_fr
most dop's don't even kno that the raw they get is not raw in the common sense
20:47
danieel
what would you do with the rawraw? if you have no access to the (proprietary) data fixing algo's from the camera/sensor makers?
20:47
danieel
it is a question of usability/purpose
20:48
philippej_fr
of course
20:48
philippej_fr
but the discussion is that dop's want to regain control of the image they produce
20:49
philippej_fr
and understand the process
20:49
danieel
most of them dont even know how to light a scene... and that is far more important than the camera
20:49
philippej_fr
those are not neutral choices
20:50
philippej_fr
danieel: that's why we are talking to real dops
20:50
danieel
and what do they need, have you asked them?
20:51
_mf
joined the channel
20:52
philippej_fr
definitely, we are doing a one week workshop in the biggest short film festival in france
20:56
danieel
so what is the most problematic thing on the other cameras from their point of view?
20:56
danieel
(except the price, for some)
20:59
philippej_fr
I have a long article to share, but its in french
20:59
philippej_fr
if you dare to google translate it, here it is : http://www.afcinema.com/La-texture-de-l-image-a-travers-le-controle-du-pique-dans-les-cameras-numeriques.html
21:00
philippej_fr
but the point of the article is that engineers don't "get" it, so good luck :-)
21:01
danieel
the table is wrong
21:01
_mf
left the channel
21:02
_mf
joined the channel
21:04
danieel
e.g. the dragon uses 19mpx for 6kFF, while for 4k it is just a crop - and can go as low as 8.3Mpx for 4kHD mode
21:05
philippej_fr
feel free to disagree with the users of our cameras :-)
21:08
philippej_fr
(denigrate would be a better word)
21:20
danieel
so what is the point, which the engineers dont get? sacrifice of DR for resolution?
21:36
philippej_fr
it's hard to summarize the long article, but in short, this dop has concerns about the lack of control he has with current cameras, and it has influence on the pictural quality of what he wants to achieve. Since we are trying to educate them about the potentialities of an open source processing, I'm looking for some explanations of what is potentionaly going on under the hood in current cameras
21:36
philippej_fr
and what we could do with this project to solve this. It's an open source project, remember ?
21:42
niemand
left the channel
21:49
_mf
left the channel
22:01
intracube
the gist of the article sounds like the DoP wants more control over the falloff of fine detail - trading sharpness for artefacts/moire/aliasing
22:05
intracube
*for less artefacts/moire..
22:33
Bertl_oO
philippej_fr: one interesting aspect could be watermarking
22:34
Bertl_oO
i.e. every camera could add an invisible watermark to the recorded images with e.g. a serial number, time, location embedded
22:34
Bertl_oO
(not in the metadata, but in the image 'raw' data instead)
22:45
philippej_fr
Bertl_oO: the guy is more asking for more control of image "quality" / processing
22:52
Bertl_oO
what I mean is, that proprietary cameras could do that already
22:53
se6astian
time for bed
22:54
se6astian
good night
22:54
se6astian
changed nick to: se6astian|away
23:08
philippej_fr
Bertl_oO: ah yes, definitely
23:11
intracube
this shows the debayer options on redcine-x at 2:40 - http://vimeo.com/47204496
23:11
intracube
the standard method is listed as 'full'
23:12
intracube
not sure if it's a fair comparison, but dcraw has several options for debayering. some are explained here: http://en.wikipedia.org/wiki/Demosaicing#Algorithms
23:14
intracube
and there was quite a noticeable difference between some of them with the raw NEF images from my Nikon
23:14
intracube
*UFraw not dcraw
23:16
intracube
giving users a range of debayer options for the beta/gamma cams would be great
23:18
philippej_fr
a good article on debayer algos http://www.stark-labs.com/craig/resources/Articles-&-Reviews/Debayering_API.pdf
23:19
intracube
it would be interesting to know what red debayering actually does
23:19
intracube
philippej_fr: thanks
23:20
philippej_fr
as they say, which is promising imho, "this is an active field of research"
23:20
intracube
so maybe this is somewhere where open solutions could give more choice than the proprietary tools?
23:20
philippej_fr
definitely
23:21
intracube
should say he's never used a red cam or software - maybe there are debayer controls elsewhere or so
23:23
philippej_fr
afaik, red provides it's own debayering (even in non red apps) and it's not user controlled
23:27
philippej_fr
arri provides it's own debayering, and a certification program for other applications, when they suit their (undocumented) "quality standards"
23:31
philippej_fr
good night everyone, and thanks for the feedback & ideas!
23:31
philippej_fr
left the channel
23:39
troy_s
All cameras are open if you are on a significant project
23:47
intracube
troy_s: in what way?
23:48
troy_s
In that any production of a given size gets a direct line to the company and the raw data
23:49
intracube
but I'm guessing there will be NDA
23:50
troy_s
Of course, but that has no practical issue