Help with hot pink artifacting in CDNGs?

Started by jonny2graham, August 06, 2013, 02:15:13 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

jonny2graham

Hey guys, new to the forum but I've been using ML+raw_rec for the production of a short I'm working on.

I'm having some problems with importing CinemaDNGs from RAWmagic/raw2dng/SonOfBatch to resolve for initial CC/LUT/meta work. On import I'm getting nasty hot pink artifacting in high contrast areas (not exclusively in the brights or darks) and I can't seem to find a workaround or fix that doesn't involve the slow/destructive RAW>DNG>ACR>TIFF>PRORES workflow. Can anyone help me out?

You can see it on the inside edges of the cardboard and especially in the water of the dam shot and all over it's brickwork/emergency release valve too :/





The workflow I'm trying to achieve, excluding audio/vfx (and which works beautifully except for this bug), is as follows:

>SonOfBatch (batch convert .RAWs to .DNG sequences, no proxies (resolve will handle that) and backup .RAWs to external HDD )
- output cDNG

>>Davinci Resolve 9 Lite (import DNGs: BMD Film log, Hunter's LUT, Initial CC and meta application)
- render ProRes 422 Proxy for editing sequences in Pr CS6
- keep Resolve project for final grade of DNGs and merge with Pr sequence XML.

>>>Premier Pro CS6 (initial editing with proxies from Resolve)
-export sequence XML

>>>>Davinci Resolve 9 Lite (final grading of masters and replacement of proxy files in XML sequence with finished masters)
-export ProRes 422 (HQ)

>>>>>>Premier Pro CS6 (final edit, titles etc)
-export final full res and 1/4 res ProRes422 (HQ)

Id really like to avoid prores for the initial stages of post (only using it once the initial CC/LUTs/crop/meta have been applied- as 422proxy for editing and as 444 later as a final graded master to swap with the proxies). I don't get any of these artifacts if I use SOB/raw2dng/QT7/Ae to output prores straight from the initial conversion but then I lose my raw handling in Resolve :(

Any help would be seriously appreciated, would help save my degree :)

jonny2graham

It's particularly bad in this take, turning the Salomon logo on the shoes pink :/ Anyone know what could be doing this?



It seems to occur between harsh dark-mid toned edges, I exaggerated it here: http://grab.by/p7Ia

araucaria

It's a know problem with resolve, you can either blur the chroma or use After Effects.

jonny2graham

Thanks araucaria, hopefully they can work out a fix for it soon :-\

jemabaris

Quote from: jonny2graham on August 06, 2013, 09:11:33 PM
It's particularly bad in this take, turning the Salomon logo on the shoes pink :/ Anyone know what could be doing this?



The best solution for this I know is the following ( I think this is the chroma blurring which has been mentioned before) : throw the clip in resolve. Under the settings tab choose "per clip" and use bmd for the color space and gamma. Pull down this nodes saturation to 0. Create a layer node. On node 3, pull down the luminance to 0.1.
Right click the mixing node and choose "add" as blending mode. Under the sharpen tab, sharpen the image until the pink disappears. Create another serial node and start grading from there.


I hope I got everything right by heart, since my PC is already shut down and I'm in bed :D good luck!

mohanohi

Resolve doesn't decode DNGs properly. Don't do any workaround with it. I use photivo to generate 16bit tiff with lowering the contrast and using Adobe ICC profile (not linear) so i could get log style image. From there you can go anywhere > Smoke > Resolve > Blender > Scratch.  This is the cleanest workflow i have come so far.
DOP, Editor, DIT, PL 5D, Ultra primes, Kowa lens

burton-fzz

the same problem with any ML raw files in resolve.
It looks like resolve couldn't debayer files from ML raw correctly. I tried RawTherapee with "amaze" debayered algorithm with 5 chunks of debayering - and got absolutely no pink dots. but (even using batch) it takes 1 frame per second even on MacPro 12 cores with 24gb ram and Mac Raid with 4 7200rpm drives and 1 system drive. I didn't tried it at home (at i5 intel processor on a home pc with windows) and can't say if it will work faster or slower... and also you losses one of main advantage - raw. if you converting with a third party debayer converter (rawtherapee, photivo) is that you got still at output. there are very good stills (16bit with 14 bit native bit depth, with better image that any camera could give at h264 or mpeg4) but they are not raw.

really good workflow is converting RAW file into DNG (of even CinemaDNG) and then import it into some adobe product and apply color balance, exposure, lens correction and color profile at importing stage and then work with native raw (dng seq).

but it don't look as pretty as in davinci resolve. =((

anyway it 50d with simple lens gives more pleasant picture in term of resolution and give more flexibility even that sony ex1r with native codec or even with convergent design nanoFlash recorder which gives only 4:2:2 8 bit image. sony workflow is really great but 50d is much more affordable and cheaper and give ability to have real post process like in hollywood cinema studios with alexa and red in hands of simple people without hundred and thousands of dollars. But it used to work really serious with material right from hitting record button until final render, and used wait hours for rendering little scene which last several seconds.

sorry if I write extra offtop but I think write something helpful =)
peace

jonny2graham

Thanks for the information everyone :) I'm working the files now in Resolve, the chrominance/luminance separation technique seems to be working well! I think my problem also may have been exaggerated by underexposure on camera- creating a lot more dark-mid edges than normal so I'm having to use quite aggressive values in the chrominance blur stage.

Yea burton I really want to work with the CDNGs to preserve the bit depth for grading so the resolve fix will have to do for now :)

This technique reminds me a lot of 'frequency separation' in photoshop for high-end stills retouching, seems to work on the same principle by separating the colour/hue information from the pixel detail/luminosity. I wonder if it is the same technique :) It's an excellent method for sharpening (without adding to colour noise).

I'll post up some screenshots of the results/settings used later, I'm off to shoot more scenes! :)
Thanks again everyone  :D

hjfilmspeed

Quote from: jonny2graham on August 07, 2013, 03:06:46 PM
Thanks for the information everyone :) I'm working the files now in Resolve, the chrominance/luminance separation technique seems to be working well! I think my problem also may have been exaggerated by underexposure on camera- creating a lot more dark-mid edges than normal so I'm having to use quite aggressive values in the chrominance blur stage.

Yea burton I really want to work with the CDNGs to preserve the bit depth for grading so the resolve fix will have to do for now :)

This technique reminds me a lot of 'frequency separation' in photoshop for high-end stills retouching, seems to work on the same principle by separating the colour/hue information from the pixel detail/luminosity. I wonder if it is the same technique :) It's an excellent method for sharpening (without adding to colour noise).

I'll post up some screenshots of the results/settings used later, I'm off to shoot more scenes! :)
Thanks again everyone  :D
This is what i have been doing. It works great. I still feel like i get better detail from ACR but im loving the speed and color of resove