12-bit (and 10-bit) RAW video development discussion

Started by d, May 22, 2013, 10:58:34 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Mayo


1%

I guess using 13 works but some calculations will be wrong?

I dunno maybe we should try 10 bit it might be less cpu intensive?

Mayo

I think what happens (regarding 13 instead of 14) is that raw2dng overwrites the wrong value anyway, so it does work in the end.

I really doubt 10-bit is less cpu intensive, but it would be nice to have as an option, for sure.

1%

I don't get it tho...  it looks like its keeping up for a while and then something happens.

Also what sucks is I can't see CPU usuage... on 600D I would be able to.

a1ex

Instead of trying to catch this unicorn (see benchmarks), I'd say it's better to investigate how to enable 12-bit output directly from DIGIC.

You can select different raw types from c0f37014: denoised, with different byte order, compressed and so on. Some test code: http://www.magiclantern.fm/forum/index.php?topic=5614.msg39696#msg39696

Mayo

Hey, don't act as if we knew this all along!  ;D

DIGIC compressed RAW... I'll try not to get overexcited this time.

scrax

Quote from: Mayo on May 24, 2013, 06:05:48 PM
DIGIC compressed RAW... I'll try not to get overexcited this time.
;D
I'm using ML2.3 for photography with:
EOS 600DML | EOS 400Dplus | EOS 5D MLbeta5- EF 100mm f/2.8 USM Macro  - EF-S 17-85mm f4-5.6 IS USM - EF 70-200mm f/4 L USM - 580EXII - OsX, PS, LR, RawTherapee, LightZone -no video experience-

a1ex

Well, some of the raw types look kinda like this: http://www.nycresistor.com/2012/08/21/ghosts-in-the-rom/

The value at c0f37014 seems to be 6-bit, not sure what each bit means.

1%

We had to test it, lol. But yea, I don't see this working right now. Forcing it lower via digic sounds like a smarter plan. Aren't those for defect detection tho?

a1ex

Some of the raws are very noisy, others are cleaned up, but there are also scrambled ones. I don't know the image format, but I know for sure the digic can output 12-bit data.

Mayo

One thing is for sure: Canon's CR2 are compressed... so it's out there.

a1ex


1%

There are several bit depths... that craw bits has several settings but for some reason they don't apply to pics or LV. I guess have to look at lv_af raw and find the address for 6D then we can start. I can probably undo the merge since this seems like a dead end unless D does something better with the module.

If we ever figure out how to use the TTJ path :)

mucher

Quote from: Mayo on May 24, 2013, 05:32:16 PM
It's just the compression ratio, it's the same for width and the whole image buffer. (to tell fwrite how much to write).

@mucher: not sure if you're serious, but all your assumptions are wrong...  :P

I know nothing about camera, nothing about programming, but only waiting too anxiously to see when this good ML stuff will come to 7D(for me a jpg sequence is perfectly fine, because it is picture style compatible--so one does not have to correct the clips' color one by one, and jpg sequence can be true color full resolution too)

1%

7D has dual procs.. no access to raw data easily. I'd do jpeg because of speed/size... a 2K jpeg would be OK... a smaller jpeg would not.

vicnaum

Quote from: 1% on May 24, 2013, 05:48:46 PM
Also what sucks is I can't see CPU usuage... on 600D I would be able to.

There's no working build for 600d? d's asm lib only for 6d?

1%

No point.. .if it doesn't run on the faster proc its not going to run on the slower one.

vicnaum

Quote from: 1% on May 24, 2013, 08:58:01 PM
No point.. .if it doesn't run on the faster proc its not going to run on the slower one.

So 100% sure it's the CPU?

Bad :(

N/A

Can we overclock the processors? I know it would probably kill the camera or something, but say I wanted an excuse to buy a mkiii...

7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

1%


savale

How about truncating the 2 lower bits? All the deep blacks have a lot of noise anyway so why not just throw them away... It will be less cpu intensive

vicnaum

How much CPU does it need more? I mean - what's the order - is it lacking 10% more CPU speed, or 500%?

1%

Truncating might be worth a shot but I dunno how to do that.

Its lacking.. ie not helping over 14 bits even though working as expected. I can't even do 960x500 continuous with this. Unless D comes back and tells us what we're doing wrong its a dead end.

bjacklee

tenacious "D" we need your help.. :D

mucher

Another maybe immature thought. 14bit color in raw video actually means 2^14 grade brightness, according to wiki.  So the 14bit to 12bit conversion seemingly should be:

int x; //brightness grade

x = x / (2^14);
x = x * (2^12);

Now things start to make sense to me.

Just my 2 coins.