When to use Tone mapping persona

[Olivier.A]

Hi,

I'm trying to better understand when using (or not) tone mapping persona, when not for merging photos. Please let me know if / when I'm wrong  I watched several videos about it but it's not that clear. But a video of James Ritson gave me some clues how it works and when I should use it.

From my understanding, Tone mapping persona is trying to slide a foot into a small shoes (sort of). I mean a unbounded (32 bits) image into a bounded space. Tools are very similar to the develop persona except the compression sliders and the clamp to SDR checkbox. Am I correct that the only moment to use the tone mapping persona is when I have a pixel layer in the photo persona in a 32 bits document (whatever the colourspace from sRGB to ROMM RGB).

When the document is 16 bits or 8 bits deep, there is no value using the tone map persona, because all tone already bounded between 0 and 1. When I say no value, I mean clamping to SDR and compression have no effect, are they ?

Also something I don't quite grab yet in that persona. Using a 32bits image, I can't see areas that would be clipped as the red area in available in the Develop persona for clipped highlights (red) and shadows (blue). May be I'm uncomfortable, because I have no visual reference of where are value above 1.. But do I need them ? (still thinking).

Any thoughts ? Thanks

[NotMyFault]

Hi Olivier,

you thoughts go into the right direction, but there is more to discover.

There are lots of different purpose to use tone map persona:

• As you explained, reduce high dynamic range from 32/16 bit color bit depth to the range which can be stored in 8-bit jpegs, see https://www.cambridgeincolour.com/tutorials/dynamic-range.htm, but combined with the next:
• To avoid loosing  too much details, it is a mapping that tries to preserve formerly visible DR by amplifying it - which can and will cause the well known artificial high contrast look of HDR images mapped to SDR DR
• Never the less, the contrast boost can be applied to (almost) any image, including SDR images which do not have high DR.
• Tone Map Persona offers options to boost global and/or local contrast

When you combine the "basic" global / local mapping ability with all the other options (contrast, brightness, shadows, highlights, curves, clarity, HSL, saturation, vibrance, B&W, ...) you get a very powerful tool to dramatically change the overall color rendering of otherwise flat looking images.

It is best practice to do tone mapping on a copy (merge visible) of the original image. This allows to play with layer opacity to de-tune the effect to taste. Editors tend to over-do the effect, so it is wise to re-visit the edit (next day or after a break).

And please keep in mind, that you perfectly calibrated PC / 12-16 bit capable ROMMRGB Display might render images in full quality (DR, color gamut etc), whereas the exported JPEG file, rendered on a different (e.g. sRGB) display with sub-optimal brightness, contrast etc. might look awful.

[Olivier.A]

Moin moin @NotMyFault,

Thank for the insight ! I'm happy to see Im in the right path. I obviously didn't know about the added contrast (I always have Trey Radcliff's images im mind). Can't this effect be done using filter layers in Photo Persona ? Or is it specific to Tone Mapping Persona ?

You suggest to us a merge visible layer because I guess Tone mapping persona is a destructive process, right ? (And that makes sense in this case)

I'm trying to keep as much data as possible until the final export, that is whyI'm trying to have the wide colourspace and tone depth possible. And yes, I notice that exporting a jpeg from a ROMM RGB, 32bits doesn't give great results. That's the reason of digging into tone mapping persona and adapt for the output.

Thanks,  continue digging !

[NotMyFault]

1 hour ago, Olivier.A said:

Can't this effect be done using filter layers in Photo Persona ? Or is it specific to Tone Mapping Persona ?

I assume both global compression and local contrast are only possible in Tone Map Persona. These functions need to access data from all (other) pixels within the image to adjust one individual pixel, possibly requiring too much cpu/gpu/ram resources to be implemented as regular filter. This is more a design choice of Affinity, not a principle limit.

1 hour ago, Olivier.A said:

You suggest to us a merge visible layer because I guess Tone mapping persona is a destructive process, right ? (And that makes sense in this case)

Exactly. You apply Tone Map Persona only to RGB layers. No RGB layer - no Tone Map.

 

1 hour ago, Olivier.A said:

that is whyI'm trying to have the wide colourspace and tone depth possible

That is a good approach in general. But it normally doesn’t make sense to stay in RGB/32 all the time:

• Some filters and blend modes are unavailable 
• The linear gamma curve could lead to unexpected results
• Currently no end user device supports 32 bit colors. You always need to convert at some point.
• 16 bit color depth is sufficient to cover all colors visible for a human eye. 

So i would recommend to do only initial color grading in RGB/32, and switch early to RGB/16 for all „fine tuning“. This will give you a more realistic rendering of how the image will look, and allow all filters, blend modes etc. to work.

It depends on you artistic (or scientific) intentions. It is waste of time and effort to optimize nuances which will be effectively invisible in the final product (export). Nice as study what is possible, or trying to discover technical limits.

Keep in mind that the process of image editing is always targeted to produce a pleasing result (in the eye of the creator or customer), and not to produce „truth“. Every edit changes the image, reducing the original „truth“ of photons that have reached the sensor. So image quality is relative. Some fine nuances of RGB/32 will get lost unavailable when converting. If you want to make them noticeable in converted exports, you need to amplify them. The only way to check if you did it correctly is to convert and inspect the result.

Sharpening is highly dependent of final image dimensions. It dos not make any sense to do final sharpening in RGB/32 and high resolution. This needs to be done as final step before export, using the final image size and a color depth not too far away from export format.

[Olivier.A]

Thank you very much for all these information and process ! They are very valuable !

I learned recently of the 3 sharpening steps (capture, creative, export) and the reason behind. I added that on my process

About the 32 / 16 process, I'm confused now. For me 32 bits means tones are not bounded, unlike in 16 bits. Colour grading in 32 bits has no much sense ? However, colour grading in ROMM RGB or Rec.2020 colourspace should be done instead of sRGB. No ?

 

[NotMyFault]

47 minutes ago, Olivier.A said:

Colour grading in 32 bits has no much sense

Nice to see someone like you really interesting into editing 👍🏼

Why do you think is has little sense? 
When you inspect the results on a system capable to render such details, it makes sense. If you intend to export to jpeg, little sense.

It might be possible someone editing in Photo, but images are used e.g. to train AI / ML models, so images never go to the display - human eye pipeline, so 32 bit images might get processed in their full beauty.

[Olivier.A]

5 hours ago, NotMyFault said:

Why do you think is has little sense? 

This sentence alone can be misinterpreted... I meant only with32 bits, color grading has no sense, because 32 bits is meant for tones. However when you combine ROMM RGB with 32 bits, then yes. First because the colourspace is large enough, so colour grading have space to find the right chroma, and adding the 32 bits depth, it also has room for the right intensity, without being stuck between 2 values (0 and 1). That's why I added ROMM RGB or Rec.2020 for colour than tones.

A good combination is a large colour pace and a large bit depth. That's why I'm trying to try to find a process with Affinity keep this combination as long as possible before exporting.

As for my background history, I was using Lightroom as my developing software, and wanted to move away. Last year I started looking around and tried Darktable. This is were I learned most of colourspace, linear etc...  (and started a youtube channel of my journey of learning). I'm far from knowing everything. That's why I'm asking questions here.