Nik color efex tonal contrast unraveled

[DanK]

Yesterday I started on an image for which I would have used Nik Tonal Contrast, so I decided to see whether I could unpack what it does. I think I came close.

First, I took the raw capture and did nothing but maximize the three contrast sliders in Tonal Contrast. I set the saturation change to zero.

Then I ported the raw capture to photoshop and made two edits, both with a luminosity blend (to separate saturation changes). First, I added a whopping amount of local contrast using USM (radius 20, amount 86). I reduced the opacity to 88%. I then added a very small amount of midtone contrast with a curve. (To be more accurate, I should have selected three ranges based on luminosity and done this separately, but I was getting lazy).

The two aren't exact matches, but they are close enough that I think I know what's going on. Here are screen captures at a fairly high magnification:

Nik Color Efex:




Photoshop:




I did a third version as well in which I did the local contrast adjustment using the texture and clarity sliders in LR because the Nik result looks a lot like clarity. This didn't work because LR works only on RGB, and increasing clarity and to a lesser extent texture increased saturation. It's clear that Nik completely separates luminosity adjustments from the saturation slider that it includes in this filter.

The next step in this for me--which unfortunately I won't get to for a while--is to try varying the USM radius. One of the differences between the texture and clarity sliders in LR/ACR is frequency: texture affects higher-frequency regions. That's preferable in many cases, IMHO, so I want to see if I can replicate that variation in USM so that I can make varied local contrast adjustments without changing saturation.

[xpatUSA]

The next step in this for me--which unfortunately I won't get to for a while--is to try varying the USM radius. One of the differences between the texture and clarity sliders in LR/ACR is frequency: texture affects higher-frequency regions. That's preferable in many cases, IMHO, so I want to see if I can replicate that variation in USM so that I can make varied local contrast adjustments without changing saturation.

An interesting post, even though I use neither NIK nor Adobe. There are probably a few ways to skin that particular cat, I reckon. I wonder if it is possible to separate out the various properties of a given image pre- and post-NIK? For example, the luminosity histograms or indeed as layers set to "Difference" or something?

I quite often use several repetitions of USM in FastStone Viewer as a pseudo-deconvolution:

First 1.3 or 1.7 radius at amount around 11.
Then 0.7 radius at amount 17.
Finally 0.3 radius at amount around 23.

It can get you some preety sharp-looking images and no reason why much bigger radii could not be used.

Also quite fond of Wavelet Contrast in RawTherapee at the various detail sizes e.g. 2, 4, 8, 16px etc. The larger sizes can bring out clouds quite well.

[DanK]

The seriously misnamed "local contrast" is much lower-frequency than that. A standard USM setting for local contrast would be a radius of 50, amount 20, threshold 0. I think this is what was recommended in the luminous landscape article that introduced local contrast via USM to many people years ago. That huge radius makes me suspect that it is akin to Adobe's clarity, which is a mixture of local contrast (that is, this sort of sharpening of low-fequency regions) and midtone contrast (that is, a curve). The texture slider omits the mid-tone contrast and supposedly sits between clarity and sharpening, affecting frequencies higher than the former and lower than the latter.

so, when I have time, I'll take a highly detailed image with a lot of texture and try USM at different radius settings. The image I posted above wouldn't be ideal--what I posted is a small crop from a very distant building captured with a 400mm lens.

On a side note, this is one more thing pushing me to more often do tonality adjustments in luminosity mode (a simpler method that is functionally equivalent to switching to Lab and adjusting tonality only in the L channel).

[xpatUSA]

On a side note, this is one more thing pushing me to more often do tonality adjustments in luminosity mode (a simpler method that is functionally equivalent to switching to Lab and adjusting tonality only in the L channel).

Can an Adobe product decompose an image into layers for example L*a*b* or HSL and then work only on the L layer and then recompose? That would address the variation in Saturation issue would it not? ...

... and Lightness is more aligned with human vision than say tonality in the form of Value.

Have you tried messing in CIECAM which allows adjustment of both Lightness (J) and Brightness (Q) in the same image? Probably not your cup of tea, though ...

[DanK]

Yes, one can switch color spaces in Adobe, make adjustments, and switch back. It's easy to switch to Lab and then back to RGB. The problem is that the adjustments within one color space are not accessible once you have switched to the other.

The advantage of using blending to do this is that all of the layers remain visible, and in fact, you can change any of them from luminosity to one of the RGB blend modes, or vice versa, whenever you want.

Some people like working on colors by working on the a and b channels in LAB. I don't, for the most part. I find it easier to work on R, G, and B. IMHO, the luminosity blend more removes any reason to use Lab unless one wants to work on colors in that space.

In any case, my main goal here was (1) to figure out what Color Efex is doing so that I can do something similar in Photoshop, and (2) to gain more control over the frequency ranges affected in local contrast adjustments. One down, one to go.

[xpatUSA]

Understood. Pardon my digging into other ways and means.

[xpatUSA]

Some people like working on colors by working on the a and b channels in LAB. I don't, for the most part. I find it easier to work on R, G, and B. IMHO, the luminosity blend [mode] removes any reason to use Lab unless one wants to work on colors in that space.

I do find LCH much more intuitive than L*a*b*.

[Manfred M]

Yes, I find LCH more intuitive than L*a*b*.

I would suggest that this is more related to workflow that any inherent issues with L*a*b*. I've spoken to several high end retouchers who work exclusively in L*a*b* and the reason for doing so.

First of all their clientele are high end commercial clients who want output for both the digital (RGB) and offset press printing (CMYK + potentially spot colours) in their final product, so an extremely wide gamut colour space lets them create files that can be cleanly converted to either. They tend to deal with high end photographers who deliver very clean digital files to them. They nail the white balance in the raw conversion stage of processing.

These retouchers spend 99% of their workflows dodging and burning, so the L* channel is is where they spend all of their time in. That gives a very fast workflow as they never have to pop out to change their blending mode to "luminosity", so the skip steps that us that work in RGB have to do. In the end they have a master file that can be converted to both an RGB or a CMYK color space very cleanly.

The products (i.e. Adobe) used in this type of work only offers these three types of colour spaces in their products. When I first started working with Photoshop, almost 20 years ago, I seem to remember that it was possible to work in LHB natively, but that was dropped many versions ago, I suspect.

[xpatUSA]

Originally Posted by xpatUSA Yes, I find LCH more intuitive than L*a*b*.

I would suggest that this is more related to workflow that any inherent issues with L*a*b*. I've spoken to several high end retouchers who work exclusively in L*a*b* and the [reasons for doing so are]:

First of all their clientele are high end commercial clients who want output for both the digital (RGB) and offset press printing (CMYK + potentially spot colours) in their final product, so an extremely wide gamut colour space lets them create files that can be cleanly converted to either. They tend to deal with high end photographers who deliver very clean digital files to them. They nail the white balance in the raw conversion stage of processing.

These retouchers spend 99% of their workflows dodging and burning, so the L* channel is is where they spend all of their time in. That gives a very fast workflow as they never have to pop out to change their blending mode to "luminosity", so [they] skip steps that [we who] work in RGB have to do. In the end they have a master file that can be converted to both an RGB or a CMYK color space very cleanly.

It would appear that neither Dan nor myself are in that "high-end" retoucher class ...

The products (i.e. Adobe) used in this type of work only offers these three types of colour spaces in their products. When I first started working with Photoshop, almost 20 years ago, I seem to remember that it was possible to work in LHB natively, but that was dropped many versions ago, I suspect.

I don't know what "LHB" is, but the GIMP offers no less than thirteen different color models to work in by means of the "Extract Channels" (decompose) mode ...

[DanK]

Does the gamut size matter? Here's what Wikipedia says about L*a*b vs. Prophoto, emphasis added:

"The ProPhoto RGB color space encompasses over 90% of possible surface colors in the CIE L*a*b* color space, and 100% of likely occurring real-world surface colors documented by Michael Pointer in 1980,..."

Yes, it certainly is in large part a matter of workflow. I don't find that changing blend modes takes much time, as even when I have a lot of dodging and burning, it's usually on only a few layers. I also find that I sometimes blend two identical tonality adjustment layers, one with a luminosity blend and one with a normal blend, when the optimal change in tonality entails some change in RGB but not as much as with a simple normal blend. And I like not losing edits done in one mode when switching to another. For me, the flexibility is worth spending a few seconds changing blend modes in an editing session that often takes quite a bit of time.

While this is also a matter of personal preference, I find it vastly easier to work on color in the RGB space.

[Manfred M]

It would appear that neither Dan nor myself are in that "high-end" retoucher class ...



I don't know what "LHB" is, but the GIMP offers no less than thirteen different color models to work in by means of the "Extract Channels" (decompose) mode ...

Fat finger typing syndrome hit: I meant HSV / HSB.

I would suggest none of us here a high end retouchers. Those folks work for the A list photographers and client organizations.

In terms of what GIMP offers, it’s fun to see what they offer and every so often something may enter the mainstream, but my feelings for GIMP haven’t changed. It’s really maintained for experimenters and hobbiests.

[xpatUSA]

Does the gamut size matter?

In my opinion, no, but there's some loose terminology in this thread.

For example, L*a*b*, CMYK, RGB are color models, not spaces, and do not have gamuts - but Manfred seemed to conflate L*a*b* with "an extremely wide gamut colour space", perhaps not intentionally.

While this is also a matter of personal preference, I find it vastly easier to work on color in the RGB space.

And "the RGB space" is a bit of a misnomer - whereas the sRGB color space is not.

[Manfred M]

but Manfred seemed to conflate L*a*b* with "an extremely wide gamut colour space", perhaps not intentionally.

In Adobe Camera Raw I can set my output to an extremely wide gamut colour space Adobe refers to as Lab Color, which has channels called Lightness, a and b.




The highly regarded color expert Dan Margulis has written at least 3 books on the subject he refers to as Photoshop LAB color.

What's a poor photographer supposed to think?


P.S. as you well know Dan and I understand what colour spaces, colour models and gamut are. Please stop bringing points up every time we don't express ourselves as clearly as you would like it.

[xpatUSA]

In Adobe Camera Raw I can set my output to an extremely wide gamut colour space Adobe refers to as Lab Color, which has channels called Lightness, a and b.

The highly regarded color expert Dan Margulis has written at least 3 books on the subject he refers to as Photoshop LAB color.

As Margulis himself says: "... but color scientists feel that we should use a more precise name for our version. They call it CIELAB or L*a*b*, both of which are a pain to pronounce and maddening typographically. Photoshop calls it “Lab color,” but the name has nothing to do with a laboratory ..."

Ref: https://biscuitsbytes.com/manuals/Margulis-Photoshop-LAB-Color.pdf

It appears to use Adobe terminology throughout the book, dumbed down for "most of us" ...

What's a poor photographer supposed to think?

I suppose you're supposed to think what the Elephant in the Room tells you to think, eh?

I, on the other hand do not so think and will continue to refer to Adobe's "LAB color" as either "CIELAB" in full or "L*a*b*".

By the way, "Lab" is not a CIE color model or space.

https://en.wikipedia.org/wiki/CIELAB...ace#Hunter_Lab

P.S. as you well know Dan and I understand what colour spaces, colour models and gamut are.

I apologize if feathers have been ruffled.

Please stop bringing points up every time we don't express ourselves as clearly as you would like it.

Oh.

[DanK]

Before getting to the pedantry, there are three questions of substance in this thread, two of which I can answer.

1. Does using a luminosity blend mode isolate luminosity from color the same way ajusting the L channel in L*a*b does? Yes, I demonstrated that some years ago here. One can obtain essentially identical results.

2. Does it matter in terms of workflow? Yes, as Manfred noted. IMHO, it's a very big difference in workflow.

3. Does the choice between working in ProPhoto and L*a*b matter in terms of output? I have no idea, but the source I quoted suggest not. I'd be interested to learn more.

Now for the pendantry:

Mea culpa. I usually type "Lab" rather than L*a*b because the asterisks are a nuisance. I suspect it hasn't caused any confusion. While that sits fine with me, I have to admit that the use of RAW in caps, which makes no grammatical sense whatever and looks like typographical shouting ("Look at me! I am so advanced that I shoot RAW!!!) grates on me, even though the corporate writers at Adobe and Canon use caps. (Well, Adobe is of two minds: you'll see that you use Adobe Camera Raw to open RAW files. Go figure.) That one does actually contribute to misunderstandings, I think.

Re referring to RGB: what I actually meant was "working in any RGB workspace", since the workflow implications of choosing not to use L*a*b are much the same regardless of whether one chooses sRGB, ProPhoto RGB, or Adobe RGB as a work space.

[xpatUSA]

Thank you for the clarification, Dan,

quote Does the choice between working in ProPhoto and L*a*b matter in terms of output? I have no idea, but the source I quoted suggest not. I'd be interested to learn more ... unquote

Me too ... anybody?

[carolus]

1. Does using a luminosity blend mode isolate luminosity from color the same way ajusting the L channel in L*a*b does? Yes, I demonstrated that some years ago here. One can obtain essentially identical results.

Is there a link to that demonstration please? I'm trying to reconcile Luminosity with Lightness and the literature is very confusing in that respect.

[DanK]

Chas,

Unfortunately, the links no longer show the images, but one of the threads is here.

If you do any kind of tonality adjustment (brightening, darkening, changing the black or white points, adding contrast with a curve, etc.) with a "normal" blend mode, you are directly changing R, G, and B values--that is, you are changing both color and lightness or luminosity. Making changes in a program like ACR or Lightroom is the same.

Sometimes, it's helpful to separate color from luminosity. For example, I often want to increase contrast without increasing saturation, and conversely, you might want to change the color values without changing lightness. There are two ways to do this in Photoshop. One is to change to the L*a*b working space, in which L controls luminosity and color is controlled by the a and b values. The other is to change the blend mode to luminosity.

The point of my old thread was to see whether the choice between those two methods matters in terms of output. The answer is that it doesn't.

Dan

[carolus]

Chas,

Unfortunately, the links no longer show the images, but one of the threads is here.

Thank you Dan - I'll take a look.

Meanwhile, I made a three-colour test image and changed it's brightness in my editor (GIMP) and the HSV colour hues and saturation did not change so I guess Adobe does it differently.

[Round Tuit]

Thank you Dan - I'll take a look.

Meanwhile, I made a three-colour test image and changed it's brightness in my editor (GIMP) and the HSV colour hues and saturation did not change so I guess Adobe does it differently.

If the test images were fully saturated to start, then changing the brightness in normal mode would not affect the saturation nor the hue.