Filmic RGB shadows

https://youtu.be/vVoIx3eMxwA?t=4250 (the whole video is worth watching, as usual)
Aurélien says filmic is not an artistic tool, and black and white in filmic are not there to manage contrast.

2 Likes

To be clear, that’s true if one uses third-party software, but if you use Canon’s own (Digital Photo Professional), you will get the OOC JPEG look right away, in higher quality. A raw file is not just the sensor data, it’s also a lot of metadata. It’s just that most software doesn’t know what to do with it.

Yes, but that’s the same situation as e.g. trying to import lightroom edits into dt: as dt doesn’t know which algorithms lightroom uses, it’s impossible to reproduce the edits completely.

Canon knows how its cameras process the raw data, so they can reproduce that processing in their software. Anyone else would either have to fork out a “certain” amount of money to Canon (if Canon is willing to sell that information…), or try to reverse-engineer the processing. And I’m not sure any camera manufacturer would want to share their processing.

he just said the same as i wrote in other words.
filmic is not an artistic tool but it’s also no tool to draw perfect curves. Also, when it comes to global tone mapping, your eyes are the best tool to decide if you are overdoing it.

From https://youtu.be/vVoIx3eMxwA?t=4348 , up to 1:13:50. He clearly says the curve should not under/overshoot.

Now that’s something that I’m curious too. It’s been in the back of my mind for some time now, but basically my question is this:

Are all raw files the same? Or better: are the raw files that come from different cameras ‘equal’ in terms of possibilities?

Let’s set aside the obvious difference that one would expect between an APS-C and a full frame sensor, or between a full-frame sensor from 10 years ago compared to a new one (say, D700 vs D810), and also the behaviour at high ISOs etc (newer sensor are expected to be less noisy at high ISO etc):
once I have a raw file from a Sony camera, one from a Nikon, one from Canon, one from Fuji, and assuming I apply the same processes to each photo, will I see a difference between each of them? A difference that can be linked back to the “Canon look” for example?.. or the “Fuji look”?

As basecurves are supposed to mimic the “look” of the different brands, couldn’t you take a few raws (playraw, raw.pixls.us, …) and apply various basecurves to get an idea? I’ve heard that Nikon uses Sony sensors in at least some of their cameras, so I’d expect at least those two brands to give similar raws. But the colours in the bayer filter matrix could be different, which would translate to (minor?) differences in colour rendering.

Caveat: Fuji sometimes does funny things with their sensors, so they might be a special case.

A little bit, because their color filters will be slightly different, which would not be a problem if they satisfied the Luther-Ives condition (i.e. if their spectral response was a linear transformation of that of the human visual system), but because they don’t, any conversion from raw space to a colorimetric space is a compromise: some spectra that would be perceived by us as distinct colors may be mapped to the same color, or vice versa. This is an interesting article on the subject: https://doi.org/10.1117/1.OE.59.11.110801

With that said, I must admit that I don’t really know the magnitude of the difference between cameras in that regard, and in any case, if you allow yourself local editing, you can always select a part of the photo and make it the color you want.

I once took shots with a Panasonic LX7 and a Nikon D7000 of the same scene, but failed to create the same outcome. Of course the two matrices may not be of the same quality, and the cameras’ dynamic ranges were quite different. I didn’t take a tripod, just hand-held them. If I find the shots, I’ll post them as play-raw, OK?

No. I don’t know the technicalities of the image data itself but just looking at the metadata with exiftool there are vendor specific data sets

Yes, of course. But do raw editors use the information in the vendor-specific (and largely undocumented) makernote section? If they don’t, then that data isn’t really relevant in comparing different raw files.

(Vendor-provided editors (e.g. Canon’s Digital Phot Professional) can of course use it, but only for the corresponding raw files.)

They need to. For instance, from my Nikon Z 6, the Black Level and the White Balance multipliers are found in the vendor Makernotes:

glenn@bena:~/Photography/rawproc$ exiftool -G DSZ_4168.NEF |grep "Black Level"
[MakerNotes]    Black Level                     : 1008 1008 1008 1008
glenn@bena:~/Photography/rawproc$ exiftool -G DSZ_4168.NEF |grep WB
[MakerNotes]    WB RB Levels                    : 1.38671875 1.734375 1 1

Edit: In my software, I get these values from libraw, the c++ library I use to read raw files. It does not specify where in the metadata it got them…

Is now completely off topic so I embrace that :).

Giml had a plugin I got from somewhere that analyzed two layers and displays a curve adjustment to go from one layer to the other.

That curve is then really, really messy and crude because it does its best to get the 2nd layer as result. But then there are options to smooth the result a bit till you see a curve that you think “so that is what they are doing!”. Save a canon raw with the neutral rawtherapee setting, or dcraw with just whitebalance. Compare that to the ooc jpeg and see what happens.

I think that all raw files are not the same,and yes I think I get what the question was about.Applying the same pipeline Canon uses on a sony sensor will not give the same results… But maybe closer.

I think most difference comes from the color filter on the sensor and the analog to digital converter.

It might be just ever so slightly more sensitive in red for example, yielding different contrast (talking minute differences here).

Or in other words, if you were (hypothetical) take the exact same picture with the exact same lens with the exact same settings… I think the ‘measurements’ in the raw file will still be (slightly?) different.

I have no hard proof though, it’s just ‘my understanding’. I might very well be wrong :).

I seriously doubt there is ‘metadata’ in raw files programs don’t use and is crucial to getting a ‘default look’. Canons own raw software just processes the data the same as the camera does. Other software has their own default rendering.

And no, achieving the same look as the ooc jpeg is absolutely not the goal for a lot of software.
(tried a lot of raw software, I never saw the same default rendering twice :)). Different algorithms yield different results.

For what it’s worth, whish my cameras had filmic + cat as the ooc jpeg look… And then don’t write jpeg but 16bit jpeg-xr or 16bit jpeg-xl. I think I could live without raw files then :).

Well, for what it’s worth, it’s a developer of LibRaw saying that: https://www.dpreview.com/forums/post/64366441

Given that the rendering depends on camera settings and other data (e.g. lens correction data), if correct interpretation of the metadata is not sufficient, it is at the very least necessary. “Canons own raw software just processes the data the same as the camera does” is precisely the point — they can do that because by reading the raw file, they have access to the same data that the camera uses.

That most software doesn’t know how to do this to the fullest extent possible (regardless of the reason) does not imply that raw files are “just” sensor data.