Current Server Time: 20:47 (Central Europe)

#apertus IRC Channel Logs

2018/02/18

Timezone: UTC


23:19
TofuLynx
left the channel
23:19
TofuLynx
joined the channel
00:38
TofuLynx
left the channel
00:44
TofuLynx
joined the channel
00:58
illwieckz
left the channel
01:02
illwieckz
joined the channel
01:03
hozer
joined the channel
01:18
davidak
left the channel
02:00
futarisIRCcloud
left the channel
02:02
aombk
left the channel
03:24
futarisIRCcloud
joined the channel
03:45
comradekingu
left the channel
08:10
Bertl_zZ
changed nick to: Bertl
08:10
Bertl
morning folks!
08:34
BAndiT1983|away
changed nick to: BAndiT1983
08:36
BAndiT1983
TofuLynx, first one has to convert the bytes, and as second step you would split the, between right layers (RGB)
08:37
BAndiT1983
you could do it in one step, but it is more cumbersome to do and error-prone
08:38
BAndiT1983
after conversion you just have to use width of the image, which is given, to calculate the row
08:50
comradekingu
joined the channel
09:44
tushar1997
left the channel
09:44
tushar1997
joined the channel
09:55
niemand
joined the channel
09:55
niemand
left the channel
09:55
niemand
joined the channel
10:16
TofuLynx
left the channel
10:16
TofuLynx
joined the channel
10:19
Guest35400
changed nick to: fredy
10:19
fredy
changed nick to: Guest7816
10:21
Guest7816
changed nick to: fredy
10:23
Bertl
off to now ... bbl
10:23
Bertl
changed nick to: Bertl_oO
10:23
tushar1997
left the channel
10:26
tushar1997
joined the channel
10:27
rton
joined the channel
10:33
TofuLynx
I think I am confused. So I load the file first using ifstream, then I load it into memory to an array of uint_8t. Then I transform each 3 bytes of the file in 2 bytes then put it into other uint_8t array, right?
10:46
BAndiT1983
yes
10:47
BAndiT1983
source array and target array, let's keep them bothe uint8_t for simplicity
10:47
BAndiT1983
*both
10:47
TofuLynx
and the for loop you suggested is for the process of converting the 3 bytes to 2?
10:47
BAndiT1983
yes, just merge 3 bytes so you can get the 2 required bytes by shifting and write them to target array
10:49
TofuLynx
that's where my confusion ensues. how do I know the target's index to save the 2 bytes?
10:49
TofuLynx
do i multiply the source's index by 2/3?
10:51
BAndiT1983
use another variable which increases by 2 in every loop iteration
10:52
BAndiT1983
imageData[j] = mageData[j + 1]
10:52
BAndiT1983
j += 2;
10:52
BAndiT1983
argh, line break got deleted, jsut a moment, will copy by hand
10:52
BAndiT1983
imageData[j] =
10:52
BAndiT1983
imageData[j + 1]
10:52
BAndiT1983
j += 2;
10:53
BAndiT1983
it's rather simple approach, but we don't need optimization currently, as many people tend to do premature optimization, first the algorithm and then one can optimize
10:53
TofuLynx
Makes sense! also in relation to reading the file into an array of uint_8t, the in.read() function only accepts arguments of type char, however I saw that I could use a reinterpret cast
10:54
BAndiT1983
you can use casting
10:54
TofuLynx
yeah thanks :)
10:54
BAndiT1983
yes, reinterpret_cast is the one, as uint8_t is unsigned char, so no need to do fancy stuff
11:16
BAndiT1983
changed nick to: BAndiT1983|away
11:44
TofuLynx
I successfully got the 4 images right :D
11:44
TofuLynx
now the debayering process
11:48
se6astian|away
changed nick to: se6astian
12:27
BAndiT1983|away
changed nick to: BAndiT1983
12:35
TofuLynx
left the channel
12:35
TofuLynx
joined the channel
12:39
TofuLynx
left the channel
13:02
TofuLynx
joined the channel
13:15
Claubits
joined the channel
13:15
Claubits
left the channel
13:16
Claubits
joined the channel
13:16
Claubits
left the channel
13:17
TofuLynx_
joined the channel
13:18
TofuLynx
left the channel
13:21
TofuLynx_
changed nick to: TofuLynx
13:28
TofuLynx
the debayering process is done, I can get color images successfully, but they have some artifacts. I'm going to try to apply some bilinear interpolation
13:33
BAndiT1983
changed nick to: BAndiT1983|away
13:35
BAndiT1983|away
changed nick to: BAndiT1983
13:41
TofuLynx
The optical low pass filter doesn't blur a little the footage?
13:55
se6astian
TofuLynx: sounds good
13:55
se6astian
there is no optical low pass filter in our camera dev kits
14:26
Kjetil
How about focus?
14:36
TofuLynx
Ah ok, I saw on the ideas list that it's possible to use the optical low pass filter. are you guys planning to use it or it's up to the user?
14:41
BAndiT1983
changed nick to: BAndiT1983|away
14:43
TofuLynx
Also I would like to get to know who you are and your roles on the project, so I can associate the nicknames to somebody :)
14:49
se6astian
currently the is no OLPF, its planned for the future, not clear when exactly
14:49
se6astian
me: https://www.apertus.org/user/7
14:50
TofuLynx
Thanks! :D
15:21
micha_
joined the channel
15:24
BAndiT1983|away
changed nick to: BAndiT1983
15:53
TofuLynx
bandit, you are responsible for the OpenCine?
15:56
TofuLynx
left the channel
15:56
TofuLynx
joined the channel
16:01
TofuLynx
left the channel
16:03
TofuLynx
joined the channel
16:04
BAndiT1983
yes
16:06
tushar1997
left the channel
16:57
micha_
left the channel
17:02
TofuLynx
It's on concept phase, what's the next phase?
17:05
BAndiT1983
what do you mean?
17:20
BAndiT1983
which page are you working on?
17:20
BAndiT1983
was meant for Rex
17:27
TofuLynx
Hmm, I mean like is there some working preview of OpenCine or its still very early?
17:27
BAndiT1983
we haven't done builds for release, but it works in several areas
17:28
BAndiT1983
just hadn't much time to continue as other apertus stuff needs attention
17:28
BAndiT1983
if you look in the lab, then you will find a couple of screenshots, which i've posted to show progress at some points
17:28
TofuLynx
Yeah
17:28
TofuLynx
I am very interested in contributing to OpenCine
17:29
TofuLynx
left the channel
17:30
TofuLynx
joined the channel
17:30
BAndiT1983
TofuLynx, you can look at uploaded files here -> https://lab.apertus.org/file/
17:31
BAndiT1983
have you already registered in Lab?
17:31
TofuLynx
Not yet
17:31
BAndiT1983
if yes, then login and try to filter for my user, there you can find screenshots, latest ones were showing progress with MagicLantern files
17:32
TofuLynx
hmm MagicLantern? Isnt it for Canon?
17:33
BAndiT1983
yes, Alex from ML was also doing stuff for apertus
17:34
Bertl_oO
ML joined us during the indigogo campaign ...
17:34
TofuLynx_
joined the channel
17:35
TofuLynx_
so ML is supporting apertus?
17:36
BAndiT1983
apertus had support from developers, at least Alex
17:36
BAndiT1983
he implemented some tools, as far as i know
17:36
TofuLynx_
what kind of tools?
17:36
TofuLynx
left the channel
17:38
TofuLynx_
changed nick to: TofuLynx
17:40
BAndiT1983
raw2dng i suppose
17:40
BAndiT1983
https://github.com/apertus-open-source-cinema/misc-tools-utilities
17:44
TofuLynx
Hmm Ah that's nice
17:44
TofuLynx
and you use that tool in OpenCine?
17:44
TofuLynx
or not at all?
17:45
BAndiT1983
nope, OC has own routines, as i'm trying to get more processing speed there
17:45
TofuLynx
Yeah
17:45
BAndiT1983
at some point processing through OpenCL, SSE/SIMD or shaders was planned
17:46
TofuLynx
so you can offload eventually to GPU?
17:46
BAndiT1983
yes, at the moment ProcessingTest is loading 3 x 16bit layers to the GPU and combines them through fragment shader
17:47
BAndiT1983
but advanced stuff like linearization, and so on, is not done yet
17:47
TofuLynx
and has it got a reasonable framerate?
17:47
BAndiT1983
we also have a de-Bayering algorithm on the list, which was developed not long ago by some people, but no one had time to implement it
17:48
TofuLynx
what debayering algorithm do you corrently use?
17:48
BAndiT1983
there were several tests with loading sequential frames, but at the moment it's just static one
17:48
BAndiT1983
simple linear interpolation
17:48
TofuLynx
ah ok
17:48
BAndiT1983
image processing was postponed, as the focus shifted to backup feature
17:48
TofuLynx
and do you think it's possible to play sequential frames in reasonable rate?
17:49
BAndiT1983
yes, but there are several factors to evaluate, and possibly to reduce resolution for preview, like AfterEffects does it by doing /2 or /4
17:49
TofuLynx
yeah, and possibly automatic resolution
17:50
BAndiT1983
while loading image data, one can take every second or 4th pixel to cut the resolution, just as an example
17:50
BAndiT1983
it would avoid the need for special downscaler
17:52
BAndiT1983
also it would reduce processing time for de-bayering, for some layers at least
17:52
BAndiT1983
usually it's not a problem to do just basic preview, but when final output would be triggered, then the application has to do full layer processing
17:53
TofuLynx
yeah
17:53
TofuLynx
Very interesting
17:55
TofuLynx
do you know some paper or site that as documented the different speeds for several debayering algorithms?
17:57
BAndiT1983
there are a lot online, don't know a specific one
17:57
BAndiT1983
but premature optimization should be avoided, if the application is really slow, then it should be considered to optimize the algorithm
17:58
TofuLynx
Makes sense
17:58
TofuLynx
Other reasons to avoid premature optimization?
17:59
BAndiT1983
many, people tend to lose themselves in this stuff, myself included, but often it is not pushing the performance
18:00
BAndiT1983
other factor like IO speed is important, as you cannot bypass that easily
18:03
BAndiT1983
TofuLynx, http://home.ifi.uio.no/paalh/publications/files/ism14-bayer.pdf
18:04
BAndiT1983
changed nick to: BAndiT1983|away
18:23
se6astian
changed nick to: se6astian|away
18:35
TofuLynx
Thanks! :)
18:36
TofuLynx
left the channel
18:45
TofuLynx
joined the channel
18:55
se6astian|away
changed nick to: se6astian
18:57
se6astian
changed nick to: se6astian|away
19:50
se6astian|away
changed nick to: se6astian
20:36
TofuLynx
left the channel
20:58
TofuLynx
joined the channel
21:05
niemand
left the channel
21:11
aombk
joined the channel
21:27
anuejn
BAndiT1983: do you think, that debayering a downsampeled image makes sense?
21:28
anuejn
wouldnt that create quite some artifacts on the edeges bc. of moiree?
21:29
TofuLynx
isn't that acceptable in previewing?
21:29
TofuLynx
also we could apply some FXAA filter, no?
21:30
anuejn
hm... i think that this shouldnt be to effective but we can experiment with this
21:31
TofuLynx
Also, can you introduce yourself? :)
21:35
anuejn
im not a core dev, just interested and building some things from time to time
21:36
TofuLynx
Ah Ok! :D
21:42
se6astian
changed nick to: se6astian|away
21:43
rton
left the channel
22:39
arpu
joined the channel