#apertus IRC Channel Logs

morning folks!

morning!

off for now ... bbl

helloo

hi supragya

except mlvfs

tried running fuse... it exits

have you looked into logs?

i sent g3gg0 i guess

hello!

hello

hi se6astian, nmdis199-

g3gg0-mobile: where in strace logs did you find the fuse exiting?

I am on mobile. The main process exited iirc just after calling mount (ret 0) and some fstat stuff

it shows no error

even on clean distro install

anyone? :) I did not receive any messages so far

shall I?

please go ahead

okay NP!

I've got an idea as to how to proceed towards it, but it definetly needs to be discussed.

Hopefully the tests should be ready to be presented really soon. That's all I guess

Yes, sure.

Currently, I am trying to focus on luma waveform , So, I am planning to read a file and then seprating the RGGB channel, once they are seperated I can use the formula for luminance which is

So, that's how I am planning to obtain luminance, Bertl suggested me today to compare the float calculations with an integer/fixed point version

sounds good, do you reduce the number of values as well eg. by averaging pixel blocks?

averaging pixel blocks is a nice idea!

A question: may I?

Yes, it is simplar and it reduced time to 0.010 on beta from 0.4s

I didn't understand? Do you mean if I am planning to plot it ?

yes

Yes, I am trying to find a way. There's an organization called kdenlive, they have successfully made a waveform tool. They used Ot library for plotting the waveform

maybe start simpler and use http://www.gnuplot.info/ ?

or gplot: http://gplot.sourceforge.net/

a luma scope looks like this: https://helpx.adobe.com/content/dam/help/en/speedgrade/using/image-analysis/_jcr_content/main-pars/image_5/lumawaveform.png

it might be useful to just show one line, one coloumn or a crop area of the image though

Hmm..I agree. I think I should go with one line/one column and then to a cropped area.

lets see where that takes you :)

Yeah :D

anything else to add? otherwise supragya would be up next

Beside, I guess supragya had something to ask?

one for color "balancing", one for contrast etc

yes, one thing after the other :)

Yep exactly

nvm...

Basic thought is to now have a sampled out LUT to translate the 12 bit RAW into appropriate color space

and for this, I had been looking over .cube and .look formats

not yet

the camera currently doesnt support 3D luts, but rather 4x1D LUTs

and the raw 12bit PLR data dumped as it is.. I guess

monitors or hdmi recorders do support 3D luts mostly though

hmm, so I am currently looking at two things... 1 is to discuss with you on how to take the LUT out (you talked about sampling etc, iirc)... g3gg0 says it should be mathematical

I thought that sampling would be better to create a LUT

Not yet

currently, I was analyzing a few CDNGs and found their LUTs have non standard behaviour

never saw one with 3x1D lut (is it possible in DNG)?

and these changes made photoshop read a DNG and not read others

(any ideas on IFD0 and SUBIFD)?

does the dng standard document contain any information about these tags?

dng have sometimes a preview image, like JPG, embedded, are you sure that you just have one data slot?

se6astian: DNG specifications just provide us with appropriate tags and give no further info... on the step size info, lut type etc etc

Opening a DNG to analyse using hexeditor is a nightmare

exiftool and specs should be enough, wrote so the loader back then

https://sno.phy.queensu.ca/~phil/exiftool/verbose.html

g3gg0, can mlvfs be changed to non FUSE solution?

That all the things for my side btw...

also, g3gg0-mobile, did you look at build logs for mlvfs? it complaint of HDutils

something to worry about>

As soon i am back from office

g3gg0-mobile: a few reference links to docs and relevant code from where to branch of would really help me

The spec defines a lot. There are various ways how to put tags into a dng and depending on the tool you use to open them, your results may vary

*stuff

thank you

Shall I?

I designed an encoded random generator using 8b10b and lfsr for the machxo2. Also a receiver which decodes and counts wrongly transmitted bits for the machxo2. Yet to perform timing analysis

great, many thanks

otherwise this concludes this weeks meeting

thanks everyone for your energy, time, patience and contributions!

with getting the PLR LUT?

or... how do you think this system that we discussed should proceed?

you can assume the light conditions to be constant

was just checking if it is feasible

(to get this LUT)

then you can vary the exposure time and plot the resulting average value against it

then you can add a PLR kneepoint and see if the response curve bends

does that sound feasible?

how do I get a lut?

yes

ah

So making one for testing purposes would be cool, but not essential

Think (damn auto correction)

I just wanted to know what are the boundaries right now... for the project per se

E.g. Swap R and G

Will try doing a few things...

(still i like your wish to know how things work)

so that maybe the image is pure magenta

Yeah or invert or swap channels

Zeroing doesnt prove that e.g. Scaling is right

it is all zeroed up

that may take some time by itself... or some synthetic frames can be generated

*broadcast color bars

I personally would start with real data

or take up mlv that is already available in samples?

Then dump that one using mlv_dump: this means MLV -> [MLV_DUMP] -> CDNG (and i have to work on modifying the MLV_DUMP?

Good day!

Ok