The CinemaDNG Discussion (raw2cdng)

Started by chmee, May 23, 2013, 10:46:55 AM

Previous topic - Next topic

0 Members and 5 Guests are viewing this topic.

JulianH

How can I determinate the correct Blacklevel? (for the 50D). Default was at 2037 but that completely crushes the file. I just tried random numbers, around 1800 gives about the same exposure as Raw2DNG, but the colors are different when opened in ACR.

Anyway, it's workable, going to find out if my computer can keep up editing cinemaDNG in resolve... Thanks for the development!

a1ex

By averaging the black areas from the DNG I've asked you to post a couple of times...

http://www.magiclantern.fm/forum/index.php?topic=5586.msg44299#msg44299

noisyboy

What a cool day! Magic Lantern community making breakthroughs all day long like an absolute BOSS  8) Amazing work!

Thanks for your hard work on this man, I can't wait to finally play with my footage in Resolve... Awesome :)

mrwolf

Thanks for the tool chmee! My footage looked good in resolve.

One question do you set the black level and white level from the raw file or is this a default setting? Should I change this value to suit the 550D? I assume this value is in the EXIF somewhere does it depend on the footage or is it always the same?

In other tools such as RAWanizer, using the tiff files, there was a purple cast and it clipped the top of the whites. While using your tool it cam out fine and I was able to access the camera raw tools in resolve which I couldn't do before.

JulianH

Quote from: a1ex on June 03, 2013, 11:48:18 PM
By averaging the black areas from the DNG I've asked you to post a couple of times...

http://www.magiclantern.fm/forum/index.php?topic=5586.msg44299#msg44299
Sorry a1ex, there is no Silent DNG option on all of the 50D builds I've been using.
I'd gladly help out, but in this case, I don't know how.

a1ex

Ask the guys to enable it in features.h.

jtkleine

@chmee, I'm shooting with 550D. I suppose it's a 14bit body.

Here's a link to a raw file:
http://kleine.me/clients/mlraw/M04-0102.RAW

Screenshots from Resove & Speedgrade:



silwerfedlt

Quote from: a1ex on June 03, 2013, 11:48:18 PM
By averaging the black areas from the DNG I've asked you to post a couple of times...

http://www.magiclantern.fm/forum/index.php?topic=5586.msg44299#msg44299

So I should do the same for my 5D2? Since the black and white levels differs from 5D3..
Or are there any precise nr for the 5D2 I should use?


a1ex

For 5D2, the black value from raw footer should be OK. I don't know if this app uses it or not.

If not, convert the raw with raw2dng and look in the exif of one DNG file.

IliasG

Quote from: chmee on June 03, 2013, 10:34:05 PM
....
@IliasG
No, logarithmic transforming is not my intention. for 12Bit i will cut LSB 2 Bit - we'll gonna see, how many data in the shadows is going to be lost - i quite think, its negligible :) Dont forget - the Magiclantern Team, all the coders around it gave you a weapon you never had the money for that - and this "techspec-more is better"- blahblah doesnt make a good film - your movie is not depending on logarithmic 16bit, but on "substance"!
...

I wouldn't have any objection for 14to12bit linear by truncating ..but on the single sample that I have seen with this method there is posterization (middle bottom on the windows frame).
http://www.magiclantern.fm/forum/index.php?topic=5601.msg39482#msg39482

Will you also truncate the "Black Level" value in exif ?.

Don't get me wrong, I am pushing this Log conversion in raw2dng as a first test to try it's effectiveness and if for example 14to12 or 10bit log will be proved of good quality ML team could use this strategy in-camera to improve the recording module for longer and/or faster fps shots ..

I can calculate and upload (within 24h) what I think are optimum LUTs (compress and linearize) if needed.   

chmee

(*) IliasG, if you're interested in implementing logarithmic scale in every kind of transcoding-process, you're welcome. (the reason why ACR didnt open the file was, the whitelevel was higher than the possible maximum output - if there was for example whitelevel 15.000 - its 12 bit overflow error). and of course the black- and whitelevel have to be corrected. so it means for me, recalculating with lin16bit->log12bit-table and embedding log->lin-LUT into cdng.

(*) i am not so sure, if recalculating in body is possible - as i understand, the possibility to record the sensoroutput in realtime is reasoned by shipping data stright to the dma, without digic and other computing-helper..

(*) jtkleine - i've seen this in my testversion of speedgrade as well - for now i have no clue, why its handling different.

regards chmee
[size=2]phreekz * blog * twitter[/size]

roughstudios

Hi everyone, fantastic work so far, big up.

I've had some time to experiment with the firmware, trying to find a workflow that works for FCPX and Resolve. raw2cdng is great but will be amazing when there's a batch version and hopefully mac also :D

Anyway, for me workflow is the thing that will decide if I'll be able to use the firmware or not, and now raw2cdng is getting close! Was trying all kinds of ways to get a cdng before but it was painful.

So now when I used raw2cdng all was good getting into Resolve, but then the issues started. First take a look at the video:



So a couple of things.

- There's a lot of what I would say is chromatic aberation. Pinkish especially in the trees. This could be a white balance thing combined with the sunset highligt, bit my best guess is that something either happened in the RAW recording or the raw2cdng conversion. Anyone with the same issues?

- Then the workflow. One word: painful. I think the timecode is breaking the connection with resolve, because there wasn't a way to get it to automatically connect to the old files. It has done in my previous tests but they were much smaller and without raw2cdng. So maybe something has gone missing in the raw2cdng conversion? All is ok until the XML is going back to Resolve from FCPX.

Any thoughts on the issues?

And keep up the fantastic work!

chmee

(*) pink fringes - it seems, its a canon problem in a strange way. other apps are handling that. it may be, there are some highcontrast-points higher than whitelevel point, and these are badly interpreted - so i have to look into another solutions, whats the best. hard limiting sensel values over whitelevel-value? and so on. (its on the todolist)

(*) timecode. hmm. could you post a xml-file in pastebin? just to see, what strange data there are? (i am win-user, so i have no fcpx)

regards chmee
[size=2]phreekz * blog * twitter[/size]

yobarry

I see a lot of people saying they want to be able to grade in Resolve with CinemaDNG, and I understand it'd be more convenient that way, but are you losing any data when you grade with converted TIFF files from ACR?

roughstudios

Quote from: yobarry on June 04, 2013, 11:42:43 PM
I see a lot of people saying they want to be able to grade in Resolve with CinemaDNG, and I understand it'd be more convenient that way, but are you losing any data when you grade with converted TIFF files from ACR?

I wouldn't say you loose that much data, it's negligible, you would loose all the benefits of not having processed the image though. Everything from white balance, noise reduction, sharpening would bring out more artefacts etc. But it would probably very subtle.

The biggest reason though is workflow, TIFF is at least 3 times as much space on your hard drive from a RAW file. If in 16 bit probably even more.

There's no way I would think that's a sustainable solution, even though at the moment TIFF is the safest workflow since it's reliable in conversion from the RAW file. Which the CinemaDNG isn't yet, but it will be!

Quote from: chmee on June 04, 2013, 10:10:50 PM
(*) timecode. hmm. could you post a xml-file in pastebin? just to see, what strange data there are? (i am win-user, so i have no fcpx)
I'll try to post both a xml with video that works with timecode and one of the broken ones tomorrow. Thanks for the great development!

yobarry

Thanks for the info

Quote from: roughstudios on June 04, 2013, 11:56:01 PM
I wouldn't say you loose that much data, it's negligible, you would loose all the benefits of not having processed the image though. Everything from white balance, noise reduction, sharpening would bring out more artefacts etc. But it would probably very subtle.
I'm a little unsure of the difference between raw and TIFF, I thought TIFF was just a container, so it just 'held' all the raw data, changing nothing?

For instance if you set a white balance in ACR and convert to a TIFF, is that white balance still 'raw' or is it baked in? And if you adjusted nothing and put sharpening/NR to zero, does the TIFF file still contain the extra data (exposure leeway) that you could have adjusted in ACR?

CaptainHook

Quote from: roughstudios on June 04, 2013, 09:13:13 PM
- There's a lot of what I would say is chromatic aberation. Pinkish especially in the trees. This could be a white balance thing combined with the sunset highligt, bit my best guess is that something either happened in the RAW recording or the raw2cdng conversion. Anyone with the same issues?

Resolve wont do any chromatic aberration fixing to cDNG's like ACR etc will. I see CA with BMCC DNG's as well. You need to fix it yourself in resolve.

roughstudios

Quote from: CaptainHook on June 05, 2013, 02:34:15 AM
Resolve wont do any chromatic aberration fixing to cDNG's like ACR etc will. I see CA with BMCC DNG's as well. You need to fix it yourself in resolve.
I know, that's why it needs to be fixed in the conversion :)

Personally I would choose a BMCC any day over the 5D if the workflow means going into manual RAW processing, then converting to CDNG to get into resolve and then edit then back for grading. It just isn't worth the time spent fixing. It might work for commercial stuff, but for documentary there's just too much material to process.

Irri

I'm also getting the pink fringes. It really shows up when adding sharpening in resolve, and goes green when using blur. It seems to be in high contrast areas. I thought it could be related to not down sampling from the raw Bayer data, but a down sample didn't seem to help.


shawnwytch

I was watching a tutorial yesterday it was covering how u can change how resolve see the footage. So it looks how u shot it, I'm at a editing job now when I get home I'll post link

EOSHD

Guys is the pink fringing the same as what I reported to chmee? Here's how I see it...



The standard Raw2DNG frames don't have this. What is that code doing differently?

sergiocamara93

Quote from: EOSHD on June 05, 2013, 11:50:41 PM
Guys is the pink fringing the same as what I reported to chmee? Here's how I see it...
I think it is. It can seem a "fringe" around contrasty areas of the image. I've tried changing to BMC Film interpretation in the Camera Raw tab and it "solves" it. But I don't really like how it re-interpret it as BMC.

Thanks a lot for this tool chmee!!! I'd love to have batch enconding and renaming but it's already great to been able to do direct RAW->CDNG processing.
5D Mark III

Irri

Here's what I get...

No sharpening (This actually looked fine until ImageShack added compression:


+5 sharpening:


-5 sharpening:


DaveAbbott

Hey guys,

I've converted a few clips to cinema dng but they're not showing up in DaVinci file browser. Any ideas?

If I use the normal raw2dng app I can see the sequence in DaVinci, but it looks like balls. If I use raw2cdng the sequence doesn't appear at all and the folder presents as though it's empty.

Thanks in advance!

Irri

Dave, have you tried setting it at 16 bit instead of 14 bit (in raw2cdng)? That worked for me.