What are your typical tone mapping methods

I am thoroughly confused. In my version 5.8 of RT, the auto match tone curve button tries to match the embedded JPEG tone somehow.

Pushing the button cancels any prior adjustments and resets the left-hand curve to Flexible and the right-hand curve to Film-like. After pushing the button the two curves can subsequently be adjusted further but I did not do that.

In other words, there are no options (unlike the 6 or so you mentioned) for the auto-matched curves other than Flexible and Film-Like.

Perhaps what I read is not what you meant?

usually you have also standard standard weighted luminance perceptual and film and maybe another one… I’m not at my PC…

Let’s try and understand one another please :heart_eyes:

1 Like

@cedric that looks better to me, no blue cast. I don’t use raw therapee so it is interesting to see your take on it.

1 Like

@Tamas_Papp i have never used the colour equalizer, i realise that I need to play with more modules and not restrict myself.

You adjust WB to taste, but I thought WB was more a technical choice and artistic tweaks are done in other modules, colour balance RGB for example. Can you elaborate a little on how you deal with WB in general, I am curious.

I take your point about seperating the bird, I am a work in progress when it comes to composition.

@Terry that’s interesting, and I take it on board. I will loosen up a bit and have a play with the other modules. I fully agree it is all about the look, in fact general composition is something I am very conscious of and working on.

I will have another stab at this and post the result along with the XMP, see if I can get more feedback.

The camera records the light as is, while a human observer’s eyes would adjust to the ambient light to a certain extent. White balance (as measured using a white surface) may not coincide with the latter, or be aesthetically pleasing. For example, most people do not ā€œcorrectā€ all the nice warm colors in a golden hour photo to neutral white, as that would deprive the image of its character.

Practically, I find Darktable’s auto WB algorithm a little bit too enthusiastic in assuming general illuminants, which is the option shown here:

screenshot_2024-12-23-085351

In this case, I usually dial back the chroma, or if I know that the shot is outdoors, go for Daylight (D) and just select a pleasing option from the slider:

screenshot_2024-12-23-085536

2 Likes

Same here. To be fair, that ā€œover- enthousiasmā€ is understandable, since dt can’t know the illuminant of the scene. So it tries to correct the selection to neutral gray. That doesn’t work too well if e.g. the scene has leaves in it which color the light: you end up with gray leaves…

SO you can either select a neutral patch (or near neutral) for the automatic correction, and if needed correct the values, or, like @Tamas_Papp suggests, pick an illuminant from the selection (for me usually daylight or blackbody) and set a decent value for ā€œtemperatureā€ (usually 3500 K – 5000 K, depending on time of day and situation). Then I adjust that ā€œto tasteā€. Note that values under 3000 K and over 7000 K are rather extreme (but not wrong per se).

3 Likes

@rvietor @Tamas_Papp Am I right in assuming you are using colour calibration to set the WB? I don’t think I am able to set the illuminant in the WB module (not at my PC right now).

I really did think the WB had to be set correctly in the WB module so that demosaic and the modules that followed all worked in a predictable way, and that adjustments to taste were done in colour balance RGB.

To an extent that would be the overall technical approach ie to get white to be white and then go from there. Much of how we assess this is strongly impacted by our brains and the perceptual nature of color both from our memory of the scene and the way we see it on the display… taking @rvietor 's comments for example

Currently DT is set to work with as-shot WB provided by the image metadata in the WB module if you use the scene-referred workflow. Then CC module will apply a CAT and try to apply an illuminant or arrive at one that will make what was neutral at the scene reflect that on your display so its not a true WB as it is not applying 3 numbers to balance white to white… Very recently instead of as shot the workflow was to set WB automatically to D65 coefficients and then run the CAT from there… You can also still use old school WB if you so choose…

The ultimate thing is the end product so don’t restrict yourself too much to a set of hard and fast rules…get a sense of how the modules respond and then use them to bend and craft an image that pleases your eyes…

If your monitor is calibrated you should end up fine most of the time…

Apparently there’s somewhat of a catch-22 with white balancing: you need a correct white balance for optimal demosaicing, but you can only set a correct white balance after mosaicing… The solution was to do a ā€œgood enoughā€ white balance before mosaicing, and then a ā€œproperā€ one later (if using both ā€œwhite balanceā€ and ā€œcolor calibrationā€, the ā€œwhite balanceā€ module is supposed to be set to ā€œcamera referenceā€ or ā€œas shot to referenceā€, if not, you get a warning.)

As @priort said, with a calibrated monitor you should get something decent. If you want an exact correction, you"ll have to shoot reference cards, and/or use known iight sources, colour temp meters, etc… Unless you have a guaranteed neutral patch in your image, WB will be a bit of a guess anyway. (and even with a neutral patch you can be surprised)

1 Like

Gents, this is really helpful. Very interesting. I do have a calibrated monitor, I use DisplayCal and a ColourMunki. I was somewhat aware of the catch 22 you mention, in that you need to WB before demosaicing.

I probably don’t want an exact balance, as you point out I would really need to apply myself far more than is practical for an amatuer just wanting to take half decent photos.

On my last trip to the zoo, when I took this photo, I did shoot a grey reference patch at different times throughout the day, just to have something to WB from as a starting point. Using this patch is new to me, something I am going to try to do more often so it is there if I need it.

I am realising that I need to not be too hung up on this, as you say it is all about the final image. If I remember right, this photo was taken at around noon, which doesn’t provide the most flattering light anyway if I understand right.

WB doesn’t come naturaly to me, I just don’t seem to have the eye for it. Often I increase the temperature slider for a heavy yellow cast, then decrease it for a heavy blue cast, then back and forth by smaller amounts until I think I see no cast.

Darktable uses the camera as-shot WB to demosaic the image (as shot to reference preset); then, you can fine-tune that in color calibration.

1 Like

Thanks Kofa. I am on version 3.8.1 which uses camera reference. I have got a lot of good info in this discussion on WB, I get the general idea of the various methods now.

This is my last edit, following this discussion. I have included the XMP and the shot of the grey card that I used for initial WB.

Grey card.
_DSC4364.nef (18.4 MB)

This was my general approach, I welcome all criticism.

. WB from grey card in the WB module.
. Disable highight reconstruction, I didn’t have any issues with strange magenta highlights.
. Denoise profiled set to wavelets auto.
. Lens correction set to defaults.
. Haze removal, I like the subtle colour boost it gives.
. Sharpen set to defaults.
. Local contrast set to defaults

. Set the exposure so the birds looked about right.
. Tone equalizer, relight:fill in preset. (played a bit with this module, and this preset seemed to lift the highlights and shadows to something reasonable)
. Set the exposure to 1.

I went back and forth between tone equalizer and exposure.

. Filmic rgb, set white and black point to taste. Played around with contrast, ended up setting it to 1. Reduced latitude quite a bit to bring back some detail in the shadows and desaturate the highlights a bit more. Set preserve chromanance to MaxRGB, I like how it desatures the blue sky.
. Colour balance rgb, auto set contrast grey fulcrum, set a small negative curve in the power slider (I like to slightly darken midtones to taste), set perceptual saturation and perceptual brilliance to try and add colour and contrast.
. Went back to WB and set it to 4300, I felt it was just ever so slightly too yellow.

. Rotate to straighten the 2 poles.
. Crop to remove the sign at the bottom right, and half the pole at bottom left.
. Export with max width of 2048 (which should also give me a small amount of free sharpening).

This file is marked with CC0 1.0 Universal. You are free to do with it whatever you wish.

1 Like

Camera reference is similar, but uses a fixed D65 illuminant as the starting point. But I think you should really upgrade to 5.0, for new features as well as for performance enhancements and bugfixes.

You might have had some bounce light too on your patch but usually shooting that gray card should give you a good starting point.

I find that you can find that sometimes CC makes the image too ā€œneutralā€ looking or just off at times but the hue is generally okay so just adjust the chroma… pulling it down relaxes the correction and increasing it will add more of a correction…For me this is usually sufficient if there is no good neutral area in the image or if using the picker on the whole image does not improve things to your liking

Sat on the couch, belly full of Christmas dinner, everyone is falling asleep and the telly is rubbish! Out with the tablet …

@kofa Agreed, it’s time I upgraded. It’s simple enough now that it is available as a Flatpak.

@priort I will play around with setting the WB in the CC module, and test adjusting chroma. In the past I just found it wasn’t quite right, couldn’t put my finger on it and no doubt it was due to finger trouble on my part.

_DSC4360.nef.xmp (11.5 KB)

1 Like

I have upgraded to v5.0 via Flatpak. Going to have a play with this image again.

(Ignoring Sigmoid) If I don’t use Filmic RGB to tone map from scene to display, at what stage in the pipeline is the image clipped to display ?

By output profile, I think. If you use a profile supporting the perceptual rendering invent, it may do something more clever than simply clipping (I don’t know; I am doubtful it handles brightness; I think the purpose of such a profile is not tone mapping, but rather colour conversion).

1 Like