Current Server Time: 07:32 (Central Europe)

#apertus IRC Channel Logs

2013/12/29

Timezone: UTC


00:12
gcolburn
okay, got the register data into the DNG and photoshop seems to not choke
00:13
gcolburn
I'll write the DNGtoRaw16 next and make sure the files are identical with diff
00:21
Bertl
excellent! thanks!
00:30
gcolburn
works!
00:31
gcolburn
I'll upload the code to github
00:34
gcolburn
I'm going to be heading out pretty soon here. If you get a chance to test the code let me know if it's working for you
00:34
gcolburn
you can use DNGtoCMV12000.py to do the conversion
00:35
gcolburn
just edit the name of the input DNG and output rawFile. I'll make it more user friendly with command line arguments later (probably tomorrow)
00:38
Bertl
great! thanks will test and report back!
00:38
gcolburn
sounds good!
00:38
gcolburn
talk to you later
00:38
gcolburn
left the channel
05:59
jucar
joined the channel
06:15
Bertl
off to bed now ... have a good one everyone!
06:44
jucar
left the channel
07:50
jucar
joined the channel
08:47
jucar
left the channel
10:14
jucar
joined the channel
10:22
jucar
left the channel
12:20
Bertl
morning everyone!
13:40
silto
joined the channel
13:50
silto
left the channel
15:20
FergusL
left the channel
15:20
FergusL
joined the channel
15:26
Celia
joined the channel
15:33
FergusL
left the channel
17:13
se6astian
joined the channel
17:17
se6astian
good evening
17:19
se6astian
hot today: chinese interest in the Axiom Alpha prototype: http://www.filmaker.cn/forum.php?mod=viewthread&tid=42322
18:00
rexbron_
joined the channel
18:00
rexbron
left the channel
18:00
troy_s
left the channel
18:02
troy_s
joined the channel
20:18
gcolburn
joined the channel
20:19
gcolburn
hello
20:23
Celia
changed nick to: FergusL
20:26
se6astian
hi Gabe
20:28
troy_s
gcolburn: Did you peek at OIIO?
20:28
troy_s
gcolburn: If you have DNG working, worth porting to it.
20:29
gcolburn
hi sebastian. working on fixing some of those issues now
20:29
gcolburn
What were you seeing with the forward matrix? just that it's there?
20:29
se6astian
great thanks
20:29
se6astian
yes
20:29
se6astian
I have no idea what it is tbh
20:29
gcolburn
whenever I removed it before photoshop couldn't read the file
20:30
gcolburn
"This tag defines a matrix that maps white balanced camera colors to XYZ D50 colors."
20:30
gcolburn
from the spe
20:30
gcolburn
c
20:30
se6astian
can you set it to all "0" ?
20:31
se6astian
or all 1
20:31
troy_s
gcolburn: http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/devnet/cinemadng/pdfs/CinemaDNG_Format_Specification_v1_1.pdf
20:31
gcolburn
identity matrix might work. I'm not sure exactly what equations the raw converter uses
20:31
troy_s
se6astian: Not useful.
20:32
troy_s
gcolburn: It is just the calibration matrix
20:32
troy_s
gcolburn: Subject to a Bradford for D50
20:32
troy_s
gcolburn: Rather PROFILE matrix. Erf.
20:33
troy_s
Idea being the RGB Bayer to XYZ, then Bradford transformed to a D50.
20:33
gcolburn
so there is overlap between tiff tags and "exif" tags. exif prefers them to be in the subIFD with the image, not IFD0. DNG allows either
20:34
gcolburn
I could try moving some tags to the subIFD and see if the exif reader works
20:36
troy_s
1.4 spec http://www.adobe.com/ca/products/photoshop/extend.displayTab2.html#resources
20:37
gcolburn
there is also an embedded Adobe profile in the DNG that when I remove photoshop would choke on
20:37
gcolburn
but if you make the calibration signatures differ it is supposed to ignore it
20:37
troy_s
se6astian: What is giving you issues?
20:38
gcolburn
hmm… I clicked that link before and got the old one for some reason. looks like that does have the new one. thanks
20:41
troy_s
How does one interpolate between two matrices?
20:41
troy_s
gcolburn: ?
20:42
troy_s
gcolburn: As I understand it there can be two matrices and it is suggested a low color temp and high (Standard A and D55-D65)
20:42
gcolburn
yeah
20:43
troy_s
gcolburn: I can't imagine they suggest LERPing between them? That is Wrong Von Kries? (Aka XYZ scaling?)
20:43
se6astian
troy_s: no issues, I just noticed some "leftover tags" in the DNG from the conversion process and wanted to make sure we have nothing unintentionally influencing our own efforts of creating a camera profile
20:44
troy_s
se6astian: The profile will be based off of RGB TIFFs methinks.
20:44
troy_s
se6astian: Pretty sure the DNG “canonized” tags won't impact the profiling attempts.
20:46
gcolburn
yeah, as long as the raw converter used doesn't use the tags
20:46
troy_s
gcolburn: The three general approaches for white balance shifting, in increasing order of psychophysical accuracy:
20:46
gcolburn
the dcraw command I've been using *should* be ignoring all tags in the DNG
20:47
troy_s
1) Wrong Von Kries (aka XYZ scaling)
20:47
troy_s
2) Von Kries
20:47
troy_s
3) Bradford
20:47
troy_s
Bradford is largely accepted as the most accurate approach as it is a matrix transform that takes the XYZ and loosely converts the data to LMS (rod / cone) response.
20:48
gcolburn
okay
20:48
gcolburn
I haven't read about the different approaches in the DNG spec. so not sure how that is handled. it might just be up to the raw converter
20:49
troy_s
gcolburn: Reading the spec it says to LERP between the two matrices
20:50
troy_s
gcolburn: How does one LERP between two 3x3 matrices?
20:50
troy_s
(for the record http://en.wikipedia.org/wiki/LMS_color_space#XYZ_to_LMS)
20:51
gcolburn
pushed fixes for the UniqueCameraModel and the 4 zero bytes to github
20:51
troy_s
And of course Lindbloom's http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html reference.
20:52
gcolburn
Sebastian: can you try the updates and see if the exif tool is happy? if not then we probably have to move tags to the subIFD
20:54
se6astian
maybe it wont influence the profiling itself but maybe the interpretation of the calibrated DNG then afterwards
20:54
se6astian
will try now, thanks
20:54
troy_s
gcolburn: Check out the ICO plugin in OIIO. About as easy I think for sample of the input / output format.
20:55
troy_s
https://github.com/OpenImageIO/oiio/tree/master/src/ico.imageio
20:59
se6astian
gcolburn, did you update the empty.dng as well or only the python scripts?
20:59
se6astian
aka do I need to run a dng through the scripts again to test?
20:59
gcolburn
yeah, that would be a good starting point to add support into OIIO. Didn't you say there is an initial patch for DNG support?
20:59
gcolburn
sebastian: yes, run a raw file through again and test on the new DNG
21:00
se6astian
ok
21:01
troy_s
gcolburn: You could likely knock out an OIIO importer / exporter in no time given your DNG experience.
21:01
se6astian
looks good
21:01
se6astian
the tags themselves
21:01
troy_s
gcolburn: And willing to bet Mr. Gritz would take the pull request.
21:01
se6astian
exif tool still cant change the tags though
21:01
se6astian
Error: [minor] DNG Version 1.4.0.0 not yet supported - img.dng
21:01
se6astian
0 image files updated
21:01
se6astian
1 files weren't updated due to errors
21:02
troy_s
gcolburn: It will make all of the code much cleaner. And much easier to iterate on the programs.
21:02
troy_s
gcolburn: Better still, it is probably easier for OCIO inclusion, which is going to be mandatory at some point.
21:02
gcolburn
so the exif tool doesn't support 1.4?
21:03
gcolburn
right now our DNGs have 1.4 encoded in them.
21:04
se6astian
seems so yes
21:04
troy_s
Junk.
21:04
troy_s
What EXIF tool is this?
21:04
troy_s
1.4 is from 2012.
21:05
troy_s
And why are we obsessing with a stills based EXIF data set?
21:06
troy_s
Great bonus sure, but god help us if people are trying to grade in a stills based EXIF system.
21:06
gcolburn
how much is Cinema DNG being used? Cinema DNG can be as simple as a sequence of DNG files with optional frame rate tags, etc.
21:07
gcolburn
so if they're reading EXIF tags for anything that may be a consideration
21:07
gcolburn
since I don't work in cinema I wasn't sure how widely that format is used right now
21:07
se6astian
troy_s, http://www.sno.phy.queensu.ca/~phil/exiftool/
21:08
se6astian
cinema dng is the only open standard raw format
21:09
troy_s
se6astian: Still EXIF is not quite motion specific.
21:10
troy_s
gcolburn: DNG had (has?) zero penetration in industry grade cameras. Only the BMCC has put it on the map regarding prosumer motion cameras.
21:10
troy_s
gcolburn: And IIRC Pentax was the only vendor using it as an option in stills bodies.
21:10
se6astian
no, but the exif tags influence the color interpretation of the DNG
21:10
troy_s
se6astian: The DNG tags do.
21:11
troy_s
se6astian: In a DNG viewer.
21:11
troy_s
se6astian: In my experience, EXIF has no bearing on color transforms in just about every stills application I have seen. Many mangle.
21:12
troy_s
(and that isn't to suggest a viewer that loads DNG _will_ mangle, but just that I would test first and trust second)
21:18
troy_s
gcolburn: Do you see there is a command line format specification?
21:18
troy_s
gcolburn: I wonder if that should be glanced at.
21:18
troy_s
gcolburn: Seems relevant to OIIO integration and the work you are doing here.
21:19
gcolburn
command line for OIIO?
21:23
troy_s
gcolburn: yes.
21:23
troy_s
gcolburn: It comes with a few tools. Many OIIO tools are based off the CLI.
21:24
troy_s
Silly reference for their CLI tool.
21:24
troy_s
Erf.
21:35
gcolburn
does sebastian or herbert have the camera now?
21:37
troy_s
se6astian Bertl ^^
21:38
troy_s
se6astian: I would encourage you to get an idealized shot done of that chart under Illuminant A (aka Tungsten ideal)
21:38
troy_s
se6astian: As well as (if possible) direct sun (no glare) in an enclosed area (to kill sky blue)
21:38
troy_s
se6astian: Because right now we have NO clue if the sensor can even reach sRGB / 709 primaries.
21:39
troy_s
se6astian: And that's _kind of_ important. ;)
21:39
troy_s
se6astian: Until we have those charts (again, captured 1) at middle grey exposure 2) 0.5 stop hot, 3) 1.0 stop hot and (ideally) 4) 1.5 stops hot) as well as evenly lit.
21:39
troy_s
se6astian: We have no idea of the profile and that means no idea of the gamut.
22:17
se6astian
I have the full grade sensor, herbert has the defect one
22:18
se6astian
I will conduct new chart shooting when the new firmware is complete
22:43
gcolburn
ok great. thanks!
22:43
gcolburn
will the new firmware have additional LUT support?
22:43
gcolburn
I could write some basic C code for the matrix LUT to Herbert and let him incorporate it in whatever way makes sense.
22:46
jucar
joined the channel
22:56
troy_s
gcolburn: *COUGH* OCIO
22:57
troy_s
gcolburn: Might as well just bite the bullet there. Dead simple to do forward and backwards matrix romps etc.
23:26
se6astian
time for bed! good night!
23:26
se6astian
left the channel
23:29
jucar
left the channel
23:29
gcolburn
good night!