An alternative method of adjusting tonal contrast & dynamic range in Rawtherapee

I asked AI:

From which I deduce that the value of Yb% is not equal to ‘Middle Gray’ except possibly at it’s default setting; and from which I deduce that to refer to background luminance as “Middle Gray” would be misleading and incorrect.

1 Like

Hello @anon8280290 , you may be right here.

Above I was just referring to a phrase of @jdc :

Mean Luminance (Yb%) is apparently the correct CieCAM term, so no need to change that to middle gray…

1 Like

@anon8280290

Thank you for these comments which confirm (for this part) the functioning of Ciecam .

But as I mentioned, this does not solve the problem of “how to calculate it ?”

I repeat what I already mentioned. I use the calculation used in DT and ART, modified by myself and weight the action by weakening the lights (fairly empirical value by default) by acting on white-distribution (+20), to try to take into account the background.

Jacques

Here’s a link to my full “conversation” with ChatGPT, where “how to calculate” was discussed … no doubt yourself, dt and ART do it differently. No criticism is intended and I am only presenting the chat for information.

https://chatgpt.com/share/7804dc30-b8d7-4da9-9372-76a8523a6ed3

You are so polite with chatgpt. I always get responses like “I’m sorry you feel that way.”


\qquad\qquad\qquad\Huge 🤭

1 Like

Sorry, I missed the question. The answer is no.

Hello

I made some additional changes to Cam16 - Selective editing (LA)

I implemented the Sigmoid algorithm (which I modified) from DT.
Now you have it also in brightness mode (Q).

As a reminder, it is currently implemented in RGB mode in the “SDA” (Source Data Adjustments) module.

Sigmoid Q, allows you to take advantage of the implementation in a CAM (Color Appearance Model) and only intervenes on the variable Q (brightness), the “rest” of the colorimetry is provided by Cam16.

It is available in “Standard” and “Advanced” mode.

It is recommended to either clear the cache or set to “neutral” to eliminate old settings.

Here in Advance mode
image

Executables
apseimprov

Jacques

1 Like

First tests using the Source data tool are very good. With sigmoid enabled and slightly raised contrast (up to 1.5) you consistently get good results using only local edit tools and no additional contrast management. Most additional tonal adjustments can be achieved using gamma,slope, midtones. Initial tests are looking very good.

Documentation on this module should mention that the standard exposure tool doesn’t work that well with this Selective Editing sigmoid. In dt and ART sigmoid makes the basic exposure tool a super tool whilst RT behaves well when the Selective Editing colour appearance tonal tools are used instead. At least my initial tests show that tones aren’t mapped/managed when using the basic exposure tool.

I’ve not yet got useful results with Sigmoid Q but that’s because I don’t know how to use it :slight_smile:

The Highlight attenuation & RGB channels label is imho a bit hard to parse but perhaps the tools are now so diverse.

1 Like

@nosle

Thank you for this evaluation. This satisfies me :wink:

One of the purposes of this topic is to suggest that there are other approaches than methods based on the “Exposure” module.
Whether it is “Abstract profile” followed by Ciecam, or here the Cam16 (Selective editing) module with Source Data Adjustments upstream.

The second case is more complete and it is difficult to find a title that can satisfy all cases.

As I said the “SDA” module is upstream of Cam16 for software reasons. But it is linked to Ciecam. Even if you do not activate anything in Cam16 Image Adjustments, Cam16 has a significant effect.

This should be seen as a whole, with here a Ciecam module simpler than that of Color Appearance & Lighting.

I’m not saying that Exposure is a bad module or that it shouldn’t be used, but it’s not the most appropriate. On the other hand, the association of TRC (gamma and slope, and midtones) with the various methods of “Highlight attenuation & RGB channels” should provide a viable solution in the majority of cases: Slope based and Sigmoid based give fairly comparable results, even if the algorithms are different. Slope based ensuring a slightly larger DR (Dynamic Range). The difference mainly depends on usage habits.

Sigmoid Q is fully integrated into Ciecam (Cam16), it takes into account the notion of “La” (Absolute luminance) compared to traditional tools which are limited and use Relative luminance. So even if your camera compensates for exposure differences between low-exposure scenes in the evening or morning and scenes in full sunlight or on snow. The absolute luminance (La) is different and Ciecam will make different corrections to take into account the physiological aspects.

Ciecam (Cam16 on Selective Editing) requires learning with the concepts of:

  • 3 process : scene (Source), Image adjustments, Viewing (display)
  • Surround
  • Various notions as Simultaneous contrast, etc.
  • Lightness, Brightness, Saturation, Chroma, Colorfullness, Hue
  • Chromatic adaptation
  • etc.

Thank you

Jacques

2 Likes

Always in Selective Editing > Color Appearance (Cam16 & JzCzHz).

I just changed in JzCzHz (Advanced mode only) the old sigmoid by the new one (close to that of Sigmoid Q)

As a reminder, JzCzHz (Jzazbz) is my interpretation of an experimental research approach concerning HDR processing as a whole.

But, “Jz” is not a CAM (Color Appearance Model). So it is my interpretations and additions that make notions like “Scene condition” or “Surround” appear, etc.

Other information, processing towards the HDR screen is not provided by this code.

image

Executables
apseimprov

Will these features, changes, etc. be in 5.11 or a later release?

@bobm
These changes will not be included in 5.11

I just made some changes to harmonize Sigmoid in Source Data Adjustments and Sigmoid Q .

There is no longer any need for “normalize”, but I added “Scale Yb scene” and “Display white point” (advanced).

@jdc

Thank you; I suspected as much.

I have been following this thread with interest but I am currently so drowned in photos that I need to process in a timely manner that I haven’t had the time to do anything more than just read.

Hello

I just made other changes to harmonize “Sigmoid Jz” (Absolute luminance Jz) with “Sigmoid RGB” in Source Data Adjustments and “Sigmoid Q” (Absolute luminance Q).

The sliders have the same name, the principle of the algorithms (sigmoid) are close, but the results differ.

Hello

I just added in “Cam16 image adjustments” the same method as in “Source Data Adjustments”.
Based on Freeman’s tone-mapping method, it uses only 1 slider in Q mode (absolute luminance).

You have the choice, either to use Sigmoid Q, or Slope based Q.
For my part, I prefer this second method - at least in this case - incorporated in the Q loop of the 6 Ciecam variables.

These methods (Sigmoid, Slope based…) complement the TRC (Tone Response Curve : gamma, slope, midtones) and not the other way around.
They are mainly used to asymptotically attenuate highlights, the added contrast is only a consequence.
For contrast, I recommend the notion of local (variable) contrast (especially with wavelets) rather than global contrast methods.

Executables, in progress
apseimprov

Good evening (in France)

Jacques

Hello

I just did a review (GUI, algorithms…) of the Tone-mappers present in Selective Editing > Color-appearance (Cam16 & JzCzHz) - Log encoding, Sigmoid Q, Slope based Q.

This should work better. This (algorithm) supposes some empirical modifications related to Ciecam, notably for Sigmoid Q and Slope based Q.

Executables - apseimprov

1 Like