Current Server Time: 11:32 (Central Europe)

#apertus IRC Channel Logs

2014/01/07

Timezone: UTC


23:00
troy_s_
joined the channel
23:02
Bertl_
joined the channel
23:02
Bertl
left the channel
23:02
troy_s
left the channel
23:10
FergusL
I expect it will end up around 10k
23:10
FergusL
well
23:10
FergusL
who am I to say this ? just a random guy
23:14
danhanes
joined the channel
00:09
danhanes
left the channel
00:16
danhanes
joined the channel
00:42
dmj_nova
joined the channel
00:43
dmj_nova
hello all, back from vacation
00:45
Bertl_
wb dmj_nova!
00:45
Bertl_
changed nick to: Bertl
00:46
Bertl
do you still remember me? :)
00:51
dmj_nova
Bertl: oh of course!
00:51
Bertl
good good ... *G*
00:51
dmj_nova
Bertl: How is the project progressing. I'm a bit behind on the ml
00:52
Bertl
well, the web pages basically show the current status, we are working behind the scenes on noise, calibration and moving pictures
00:54
Bertl
probably the greatest change after the boxed alpha was the new FPGA code (train2/snap2/hist2)
00:54
Bertl
which finally got rid of the lvds issues
00:55
Bertl
sebastian is busy mostly doing test images and charts with the boxed alpha
00:55
Bertl
+grammar
00:55
dmj_nova
very nice
01:10
Rambutan2000
Good Morning from Singapore!
01:10
Bertl
morning Rambutan2000!
01:30
Bertl
Rambutan2000: what's new in singapore?
01:31
Rambutan2000
Well it's still hot :p
01:35
Bertl
yeah, like 30°C I guess, no?
02:17
danhanes
left the channel
03:04
ThatCantBe
left the channel
03:21
jucar
joined the channel
04:31
aPinky
left the channel
04:40
Topic
apertus° - open source cinema | www.apertus.org | Currently Hot: Axiom Alpha Development http://axiom.apertus.org/ | IRC Logs available at: http://irc.apertus.org
04:40
se6astian
has set the topic
05:51
Bertl
off to bed now ... have a good one everyone!
06:46
se6astian
joined the channel
06:51
se6astian
good morning
07:14
PhilippeJ
joined the channel
08:42
PhilippeJ
Black betty making of story http://blackbettycameras.com/building-betty-2/
08:58
Rambutan2000
Sorry haven't been on much, everyone has come back from holidays at work this week so a bit busy, I've been think of some design stuff, se6astian I'll PM you when I've got something to show
08:59
se6astian
Rambutan2000, sounds good, looking forward to it!
09:26
troy_s
joined the channel
09:26
troy_s_
left the channel
10:00
PhilippeJ
left the channel
10:46
PhilippeJ
joined the channel
11:22
PhilippeJ
left the channel
12:07
PhilippeJ
joined the channel
14:11
dmj_nova
left the channel
14:32
dmj_nova
joined the channel
15:02
dmj_nova
left the channel
15:02
dmj_nova
joined the channel
15:11
jucar
joined the channel
15:25
jucar
left the channel
15:32
Bertl
morning everyone!
15:32
dmj_nova
left the channel
15:33
dmj_nova
joined the channel
15:36
se6astian
hello!
15:46
PhilippeJ
left the channel
16:03
dmj_nova
left the channel
16:03
dmj_nova
joined the channel
16:31
dmj_nova
left the channel
16:31
dmj_nova
joined the channel
16:32
dmj_nova1
joined the channel
16:32
dmj_nova
left the channel
16:32
dmj_nova1
left the channel
16:32
dmj_nova
joined the channel
17:34
dmj_nova
left the channel
17:34
dmj_nova
joined the channel
17:35
dmj_nova
left the channel
17:35
dmj_nova1
joined the channel
17:36
dmj_nova1
left the channel
17:36
dmj_nova
joined the channel
17:54
Bertl
dmj_nova: testing your IRC client?
17:58
dmj_nova
Bertl: not intentionally :P
18:58
dmj_nova
left the channel
18:58
dmj_nova
joined the channel
20:18
gcolburn
joined the channel
20:18
gcolburn
hello
20:19
Bertl
hey gcolburn!
20:19
gcolburn
how's it going?
20:19
Bertl
slowly but steady :)
20:19
gcolburn
yeah :0
20:19
gcolburn
i've been a bit busy lately
20:19
gcolburn
but thought I'd see if I can get a calibration to work
20:20
Bertl
excellent, if you need help with the geometric corrections, just let me know
20:20
gcolburn
yes, please. I don't know if I have any programs installed right now that can do that
20:20
gcolburn
without possibly modifying the images in other ways
20:21
Bertl
okay, give me the md5sum of the raw you need corrected
20:22
gcolburn
MD5 (it8-tungsten-01-12ms.raw16) = 9ad01d2150b87293cdb4feea7b2b06c0
20:23
Bertl
okay, got it
20:23
gcolburn
what tool are you planning on using?
20:23
Bertl
imagemagick
20:23
gcolburn
ok
20:24
Bertl
I'll give you the commands once I'm done
20:24
gcolburn
okay. I don't have it on my mac at the moment
20:25
Bertl
np, will upload the corrected file as 16bit rgb png?
20:26
gcolburn
sure
20:34
gcolburn
hmm⦠guess perspective correction is an option that wasn't turned on in Argyll⦠I'm trying it now
20:34
gcolburn
hmm⦠it made it past that step now
20:36
gcolburn
average error is now 12.5%
20:36
Bertl
okay, no problem, I'm almost done, will upload the transformed version as well
20:36
gcolburn
sure
20:36
gcolburn
we can compare
20:44
gcolburn
I tried running the profiling on the chart with no rotation/cropping and the mean error was 40%
20:46
jucar
joined the channel
20:48
Bertl
still no idea what those percentages actually mean
20:51
gcolburn
yeah
20:51
gcolburn
:)
20:51
Bertl
I'm 200% sure that only 50% of the 75% average is correct :)
20:51
gcolburn
i would guess that once it generates the matrix it applies each patch through the matrix and computes an error from the expected XYZ values
20:53
gcolburn
hmm⦠this time I tried applying a median filter in DCRAW (suggested by troy), and the relative error is 21%
20:54
gcolburn
did you see that Truesense imaging provides calibration matrices for each sensor? Obviously that doesn't account for differences between individual sensors or batches, but I would be their calibrations are more advanced then ours
20:57
Bertl
yup
20:57
Bertl
'ours' being 'cmosis' :)
20:58
gcolburn
well, I was referring to us shooting an IT8 chart ourselves and running through Argyll. I would hope the manufacturer has a better method :)
21:00
Bertl
yes? .. I'm not convinced they do/have/care :)
21:02
gcolburn
yeah
21:02
gcolburn
I think Truesense might care more than CMOSIS
21:02
gcolburn
they said they have a dedicated color scientist that we could consult with
21:03
Bertl
yes, sounds interesting
21:03
gcolburn
and I believe they have more sensors used in photography than CMOSIS
21:03
gcolburn
(the medium format CCDs)
21:03
gcolburn
by the way, I've been communicating with Illunis Cameras
21:04
gcolburn
they said instead of camera link and GigE they can provide a ribbon cable that has the output of their processing that they would feed to the Camera Link chip
21:05
gcolburn
its an LVTTL signal with frame/line/data/pixel valid clocks, a serial port for controlling the sensor, and a trigger port
21:06
gcolburn
they expressed interest in the zed board (they hadn't seen it before I guess), and were wondering if it could be used to record to an Sata SSD
21:07
Bertl
processing being what exactly? and what are the data rates?
21:07
gcolburn
well they have a lot of built in options like LUTS, FPN reduction, etc.
21:07
gcolburn
their cameras have 3 boards
21:08
gcolburn
sensor board, fpga/microprocessor board, power/communications board
21:08
gcolburn
these are mainly for CCD sensors
21:09
gcolburn
full frame or medium format
21:09
gcolburn
so the larger sensors have slow frame rates
21:09
gcolburn
2-8 frames/second
21:10
gcolburn
they have nikon f-mount as well as medium format mounts
21:10
se6astian
left the channel
21:10
gcolburn
while looking in their user guide I saw control for canon lenses as well (I presume they reverse engineered it)
21:11
gcolburn
http://www.illunis.com/support/assets/RMV%20Manual%20Rev%20EA.pdf
21:12
Bertl
uploading corrected images now, will take a little
21:12
gcolburn
okay
21:12
gcolburn
thanks
21:13
Bertl
I've made 4 different version, I'd be interested in the differences they produce
21:14
gcolburn
definitely
21:16
gcolburn
http://www.illunis.com/oem_custom/oem_custom.html
21:16
gcolburn
you can see how the 3 boards are laid out in a fairly small package
21:17
Bertl
yeah, I saw that page last time
21:18
gcolburn
ah ok
21:19
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/it8-tungsten-01-12ms-corr-des.png
21:19
Bertl
that's the first one
21:20
Bertl
let me know if that is sufficient for the tools to find the patches
21:20
gcolburn
ok
21:20
gcolburn
i'll try it now
21:20
Bertl
if not, I can also do a cubic correction
21:28
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/it8-tungsten-01-12ms-corr-enh.png
21:30
gcolburn
15 % error
21:31
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/it8-tungsten-01-12ms-corr-med.png
21:32
gcolburn
when I profiled the image from sebastians canon I got 3.67%
21:33
gcolburn
I'll try a few more images, but I don't think the perspective/crop/rotate will be able to get us that low
21:34
gcolburn
what's the difference in the last two images? (currently downloading)
21:35
Bertl
I did some post processing on them which should not affect the calibration but might improve on the noise
21:40
gcolburn
okay
21:40
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/it8-tungsten-01-12ms-corr.png
21:40
Bertl
that's the last and unprocessed one
21:42
gcolburn
ok, thanks
21:43
Bertl
I would expect the lowest 'noise' in the -med version
21:56
gcolburn
I just tried using the matrix with the 12% error in photoshop raw
21:57
gcolburn
compared with the previous matrix that was 20%
21:57
gcolburn
there is a slight white balance difference
21:57
gcolburn
but when I white balance both using the same gray path. the images are indistinguishable
21:57
gcolburn
path->patch
21:58
gcolburn
and those matrices were created using different images captured at different times
22:02
gcolburn
I get 15% error with the "med"
22:02
Bertl
so the same as with the enh, yes?
22:02
Bertl
can you also run the unprocessed for reference?
22:03
gcolburn
i've only done the des and med so far
22:03
gcolburn
the unprocessed that I made using DCRAW was 21%
22:04
gcolburn
enh is 15% as well
22:05
Bertl
okay, if the unprocessed gives 15% too, it's just based on geometry
22:06
gcolburn
could be
22:06
gcolburn
i'm still curious why the canon was so much better :)
22:06
gcolburn
if it was just that he could get a better framing of the chart, or if its that combined with noise
22:07
gcolburn
even though our noise reduction techniques haven't improved anything thus far
22:07
gcolburn
I can still see FPN in the images
22:07
Bertl
yes, nothing was removed
22:08
gcolburn
so, I guess the question is where to go from here
22:08
Bertl
all post processing is generic stuff
22:08
gcolburn
I think for now the calibration is probably decent, since differences in mean error isn't visually changing anything
22:08
Bertl
well, I think we should create a perfect chart and see if that gives a perfect score :)
22:15
dmj_nova
left the channel
22:34
danhanes
joined the channel
22:34
Bertl
wb danhanes!
22:34
danhanes
Hi Bertl - Thanks. Just looking at the log & your discussion of sensors
22:35
danhanes
lots to do
22:37
jucar
left the channel
22:37
Bertl
yeah, we are counting on you taking over the FPGA development soon :)
22:38
gcolburn
sorry I had to step out
22:39
danhanes
Ha - I just hope I can help in some way
22:39
gcolburn
danhanes: Hello. I don't believe we've met. My name is Gabe
22:40
danhanes
hi gabe - I'n new in these parts. Hoping to come up to speed on the FPGA development
22:42
gcolburn
great. I imagine Herbert would appreciate help with that :)
22:43
gcolburn
I'd like to learn more of the FPGA side but I've been working primarily on the sensor characterization
22:43
danhanes
I am astonished at what he has accomplished. I did some VHDL a few years ago, and it the skill is slow in returning
22:44
gcolburn
yeah, I've only done a little bit of VHDL.
22:44
danhanes
I was looking at your discussion just now - is the choice of the CMOSIS part being reconsidered ?
22:44
gcolburn
no
22:45
gcolburn
not that i'm aware of
22:45
gcolburn
I'm working on a separate project that Herbert is helping me with
22:45
gcolburn
I'm looking into the Truesense Imaging sensors (formerly Kodak)
22:45
danhanes
OK. I see. I was looking at the CMOSIS part for a project about a year ago - thats actually how I came upon Apertus.
22:46
danhanes
Much better to collaborate
22:46
gcolburn
same here
22:46
gcolburn
haha
22:46
gcolburn
so we're trying to find mutual areas of interest
22:48
danhanes
Kodak had (has) some good CCD tech, but I haven't followed their CMOS parts. Don't know much about the Truesense spinoff
22:50
danhanes
Bertl: I hate to keep asking background questions, but I will anyway....:)
22:51
Bertl
no problem, go ahead ...
22:51
danhanes
is the goal for the alpha prototype to output only HDMI at 1080 (4:4:4) ?
22:52
danhanes
then a hw redesign ?
22:52
Bertl
for the live output, yes, full HD @ 60FPS will be the max
22:56
danhanes
I understand that the immediate need is for a functional prototype prior to crowdfunding. Is there a target spec, or a time-derived requirement ?
22:59
Bertl
well, no need, but we want to make sure that we can pull it off as planned