CinemaDNG highlight recovery in Resolve vs. Photoshop

I'm finding it hard to recover highlight detail from BMPCC CinemaDNG files in Resolve, but in Photoshop -- using Adobe Camera Raw -- the highlight detail is recoverable.

Attached is an example from a portion of a BMPCC image processed in both programs. The only adjustments made were to reduce exposure in ACR and both Exposure and Highlights in Resolve Camera Raw. Much of the highlight detail Photoshop/ACR was able to recover is just not recoverable in Resolve.

No matter what combination of Camera Raw settings I try in Resolve, I can't replicate my Photoshop/ACR results -- ACR does it better. Additional adjustments in Resolve don't get me any closer either. The workflow in Resolve is much, much easier, however, and I want to figure out how to get the same results there that I do in Photoshop.

Any ideas on what I'm doing wrong or what I can change to get better results?

not to be too obvious but did you remember to click the 'recover highlights' button in the RAW menu?

[Rohit Gupta]

Could you post a DNG frame so we can analyze it? I'm assuming you did enable the highlight recovery checkbox in Resolve.

It also depends how you process footage in Resolve. Different LUTs and different color and gamma transforms produce different results. There is no common look for RAW files in Resolve same as in ACR.

For the posted image I didn't enable Highlight Recovery because it turned the clipped portion of the highlights pink -- it didn't recover any detail.

It also depends how you process footage in Resolve. Different LUTs and different color and gamma transforms produce different results. There is no common look for RAW files in Resolve same as in ACR.

Before applying LUTs or any other processing, my goal is to get the exposure corrected in Camera Raw -- same as one does in Photoshop: first ACR, then additional processing. So at this first stage, no other changes were made to the image. I know the detail exists in the DNG since ACR is able to draw it out -- but I can't make Resolve do the same.

I'll try to post a DNG file later to see if someone can get the ACR recovery results with Resolve.

Here's a link to an overexposed CinemaDNG file. Below is a JPG of the same image processed in Adobe Camera Raw. I can't replicate these results in Resolve. Maybe someone else can and could give some pointers on how to do so.

Yep, pink highlights in Resolve. 2 stops overexposed footage. Pocket. This is interesting. Seems ACR adds some deeper processing of extremely clipped highlights. May be some kind of artificial intelligent algorithm or better DNG support? Or not too perfect Pocket color science?
I can only suggest don't overexpose things and image will look great.

It's possible to work in it in Resolve and get acceptable results. But there is a point where these cameras get overloaded.

[image deleted by request]

It's possible to work in it in Resolve and get acceptable results. But there is a point where these cameras get overloaded.

Yes, but I think his question is valid -- why doe Adobe ACR handle these clipped highlights better than the Resolve recover highlights mode?

Clipped highlights looking pink mean you have clipped one or two channels, likely green and blue. The colour Resolve reports is probably technically more honest, but not what you want. Every raw or PreRes shoot I do uses (as applicable):
In camera:
- Film mode, not Video,
-12 bit recording e.g. 444/XQ,
In Resolve;
- soft clipping,
- highlight recovery, and
- 10bit full.

If you're having to deal with available light with very bright areas, either bring the light down sufficiently using 85-90% zebras to ensure all colour channels have detail or let it clip completely and control the look of the clipped area in post.



Sent from my iPhone using Tapatalk

[Dmitry Kitsov BMD]

An interesting strategy of dealing with this is to use the highlight recovery to maximize number of the pixels recovered, without limiting them by the values of other channels or averaging the recovered pixels. That way all of the pixels that are possible to recover are recovered, but because clipping happens to a different degree for different channels you will not achieve grey scale balance in most cases. Certainly, if you know that a particular object that got clipped was very neutral an easy approach would be simply mixing all of the pixel values from all three channels for restored highlights together, achieving neutral monochrome texture. Arguably this would work for 90 % of cases, especially when the clipped highlight is assumed to be a specular (one cannot be helped not to clip those) or something like a white or grey t-shirt.

This would not be a valid approach when the highlights clipped are not neutrally colored objects, or when the light is assumed not to be neutral white.
Manual approach is available. One can qualify by limiting to the restored highlights only, and use Curves Lum Vs Sat. This will work well for those clipped highlights that are intended to be neutral.

You may, again, qualify only the restored clipped highlights and then use RGB mixer to dial in the influence of restored pixels from different channels on output pixels of other channels. This will allow for a finer degree of control and is invaluable when color of the highlight areas clipped is not neutral on the object in the scene.

the implementations of highlight recovery/color reconstruction look quite unsatisfying in most video RAW solutions. they are much inferior in comparison to state of the art photo RAW processing.

e.g.:
https://www.darktable.org/usermanual/ch ... nstruction
https://www.darktable.org/2015/03/color-reconstruction/
https://www.darktable.org/2016/10/raw-overexposed/

It's possible to work in it in Resolve and get acceptable results. But there is a point where these cameras get overloaded.

This looks really good to me, Marc. Would you be willing to share your process? How did you reveal so much highlight detail while getting rid of the pink color cast?

On the Color tab, Camera Raw panel:
You will find just about all the existing Camera Raw settings in Photoshop.
Do not keep the parameters auto and switch to manual: Decode Quality, Decode Using CLIP, etc ..:
After that you can use all the other parameters

(Another image variant exclusively using the Camera Raw panel with the Highlight Recovery ON)

On the Color tab, Camera Raw panel:
You will find just about all the existing Camera Raw settings in Photoshop.
Do not keep the parameters auto and switch to manual: Decode Quality, Decode Using CLIP, etc ..:
After that you can use all the other parameters

I just can't replicate anything close to those results in terms of highlight recovery. I've set quality to Full, Decode to Clip, Blackmagic color space, Gamma blackmagic design film, highlight recovery checked -- and even taking highlights and exposure all the way down (just experimentally) I still have almost no highlight detail. Plus the highlights go totally pink.

Would you be willing to share your exact Camera Raw settings, Jean Claude? A screenshot, perhaps? Are you doing anything with Curves?

If they go pink, just desaturate the highlights - easy fix.

Adobe is likely using an process that analyses the color of the non-clipped areas surrounding the clipped areas. They have improved the process over the years, earlier it used to grey out clipped colors. But that process might be computationally too intensive to run in realtime on video.

On the Color tab, Camera Raw panel:
You will find just about all the existing Camera Raw settings in Photoshop.
Do not keep the parameters auto and switch to manual: Decode Quality, Decode Using CLIP, etc ..:
After that you can use all the other parameters

I just can't replicate anything close to those results in terms of highlight recovery. I've set quality to Full, Decode to Clip, Blackmagic color space, Gamma blackmagic design film, highlight recovery checked -- and even taking highlights and exposure all the way down (just experimentally) I still have almost no highlight detail. Plus the highlights go totally pink.

Would you be willing to share your exact Camera Raw settings, Jean Claude? A screenshot, perhaps? Are you doing anything with Curves?

Hi denny

Can you post a link for the source DNG: I can not find the link anymore.

Thank you

This looks really good to me, Marc. Would you be willing to share your process? How did you reveal so much highlight detail while getting rid of the pink color cast?

Not to be mysterious, but I'm reluctant to share secrets for free. I can tell you that I take the advice from the Disney film Zootopia and try everything. When you've done it long enough, eventually you can do this very quickly.

I can tell you that Dimitri from BMD has lots of good advice above. The RGB Mixer, Highlight Recovery, and the Highlights control all can help when used in tandem. There's a lot of situations like this where there is no one right answer -- it's often two or three things you have to use at the same time, in a specific order. Trial and error will eventually give you the correct results.

I think MixingLight.com had some good insights awhile back on how to recover over-exposed footage, and there's some very good ideas there. (Be aware this is a pay site.)

There's a lot of situations like this where there is no one right answer -- it's often two or three things you have to use at the same time, in a specific order. Trial and error will eventually give you the correct results.

you are right! -- and that's exactly the reason, why a single "highlight recovery"-checkbox isn't a sufficient solution in this case! there is no single way to solve highlight recovery in all cases automaticly. giving the user more control to choose beetwen a few differtent well known strategies/algorithm to handle this task is much more adequate way to solve this problem.

but it's also a very interesting example for a more general class of problems: highlight recovery/color reconstruction should be handled before debayering for optimal results. but modifications to this stage of the processing pipeline aren't available to OpenFX plugins or similar third party extension. that's very unsatisfying, because nobody can improve it by porting existing better solutions to davinci.

There is no magic bullet but I find that what exists in the camera raw panel of Davinci Resolve allows many things identical to what is found in that of Adobe camera Raw of Photoshop, which was what requested OP.

Between working, ISO, exposure, lift, contrast, etc ... allows to get a fairly correct result to recover some Fstop.

I posted something a little over a year ago regarding this particular subject. Basically it's a way of automatically getting the extra highlight information so that you can grade it directly (therefore bypassing the need for the highlight recovery function).

http://www.magiclantern.fm/forum/index.php?topic=15801.msg164822#msg164822

Bump for continuing relevance...

As the Original Poster found, I have ALWAYS been able to bring back 15% - 20% more extreme highlights in Adobe Camera Raw than in Resolve. And yes, this makes zero sense, but it has been the case for years and years, and has never been truly addressed in Resolve.

Bottom line is, if you have a shot in RAW with blown highlights, first process it in After Effects, or Photoshop or Lightroom, and there is a great chance you can save that shot, and you'd never know this if you just used Resolve.

Recovered shot

Screenshot 2023-06-26 at 20.59.35.jpg (735.18 KiB) Viewed 891 times

Original shot

Screenshot 2023-06-26 at 20.58.53.jpg (724.96 KiB) Viewed 891 times

I am a total novice by comparison the group but would like to share an observation: I think Resolved DID have the same ability to recover blown highlights but something changed when it transitioned to Braw. I have been able to recover footage that looked entirely white during one of my surgeries. It was shot on a BMPC 2k. I randomly examined a frame in Lightroom and miraculously everything recovered. With the tools discussed here Resolve did the same. I have since tried to replicate this with newer cameras and in more recent versions of Resolve to see if I can record intra-op footage that has very challenging lighting differences and it always showed clipping.
Has anyone tried to see if Premiere has the same abilities as Photoshop and Lightroom for recovering overexposed footage?

Apologies- I have realised that the Resolve has similar abilities to Photoshop and Lightroom WITHIN some parameters:
- I no longer own a BMPCC so I filmed heavily overexposed footage on Nikon Z9 in ProRes Raw and Nikon Raw , both in N-log.
- I converted the Prores Raw footage to CDNG with RawConvertor
- I set the Resolve project to DaVinci YRGB Colour Managed as advised in a Youtube video for Nikon Z9
- In the Camera raw settings I selected Decode using : Clip
- From the same section I used the Exposure settings and ... NO MORE CLIPPING . Snowstorm footage to perfectly usable footage was again possible.

This worked similarly with Nikon Raw footage.

No Highlight Recovery was used.

My assumptions are therefore:
- Only uncompressed raw footage retains the necessary data to recover significantly overexposed footage
- The same may be true of BRAW at maximum quality in the film mode rather than video mode.

Any previous attempts without log and with some compression showed clipping when I tried in BRAW , Nikon Raw and ProRes Raw. I have never tested this on maximum quality BRAW.

I suspect a lot of the mathematics behind these transforms are not shared with the users ( and for my part I wouldn't understand them) but the "un-necessary" data may be promptly discarded when there is compression or LUT application.