Contrast adjustment in RGB mode saturation increase

[digitaldog]

Unnatural looking edits, not an act against nature.

IMHO, the current behavior you're against produces far more natural looking edits to the alternative.
And of course there are the tens of thousands of ACR/LR users and their massive number of processed images that highly suggest there's not a problem, and that these thousands of customers have no issue whatsoever with how Thomas handled this behavior.

[Peter_DL]

IMHO, the current behavior you're against produces far more natural looking edits to the alternative.

But then Eric Chan’s comment (which you kindly provided in post #4) clearly suggests that there is room for improvement.

--

Eric Chan wrote:

Thomas and I separately explored the side effects and implications of various tone curve implementations (he in the early days of Camera Raw, I when working on color profiles in 2008 and again in 2010).  I learned a few things along the way.

One is that contrast and saturation are often correlated in the real world, which is why an increase in contrast often (though not always) works well with an increase in saturation.  Clarity is an example of a contrast control that doesn’t obey this:  it allows you to increase/decrease (local) contrast while preserving saturation.  If you’ve ever tried a strong negative Clarity adjustment on a colorful image, the result looks a bit odd.  Similarly, if you’ve taken a rather pale image and added strong positive Clarity but without punching up the Saturation/Vibrance controls, the result also tends to be artificial.  So we’ve not been entirely consistent in how we’ve dealt with saturation side effects in our contrast controls, but I’d probably give the nod to Thomas’s design (saturation side effect).

A second lesson is that the choice of color space makes a big difference to the result.  As Jeff, Andrew, and others have widely documented, we use ProPhoto as our choice of RGB primaries, which is a good thing in many cases.  For saturation side effects in curves, though, there are some issues.  In particular, due to the position of the blue ProPhoto primary, this means our current tone curve tends to have much stronger blue saturation side effects than in other hues.  In particular, if you have a typical S curve or just increase the Contrast control, darker blue tones (such as first column, third row of a standard ColorChecker, or deep blues in water reflections) tend to get overly saturated, and lighter blue tones (such as skies) tend to become overly desaturated.  There are ways to get the saturation side effects to be more perceptually uniform, and I’m investigating those for the future (would likely require a process version bump, though).

A third lesson is that our current color control set within the ACR/LR UI isn’t really good (yet) for doing 3D color edits.  This is because one cannot fully control how hue and saturation are affected as a function of lightness (or brightness, or luminance, or whatever term you want to use).  For example, when you use HSL controls to change the Orange hues, that changes them for all orange hues, light and dark.  You can’t change them separately for light vs dark hues.  It is possible, of course, to use per-channel RGB curves to bring back Photoshop-style hue twists in tone curves, but even though you don’t have full control over how those twists behave.

[digitaldog]

But then Eric Chan’s comment (which you kindly provided in post #4) clearly suggests that there is room for improvement.

There's always room for improvement like providing the user both options. Assuming again:
1. There's an outcry from users that they must have it (there isn't).
2. It's not going to complicate the product for the gain of a tiny few users. 

[Tim Lookingbill]

But then Eric Chan’s comment (which you kindly provided in post #4) clearly suggests that there is room for improvement.

In particular, if you have a typical S curve or just increase the Contrast control, darker blue tones (such as first column, third row of a standard ColorChecker, or deep blues in water reflections) tend to get overly saturated, and lighter blue tones (such as skies) tend to become overly desaturated.  There are ways to get the saturation side effects to be more perceptually uniform, and I’m investigating those for the future (would likely require a process version bump, though).

Simple fix for darker saturated blues requiring no process version bump IMO. I do this all the time, no complaints. Adjustment brush works just as good. I can quickly add color to change the blue using its color picker. This A/B sample below was done in ACR 6.7 which means this was fixable a long time ago.

[Bart_van_der_Wolf]

...in your opinion. So far, your arguments fall back to numbers and not appearance. The video you alluded to failed to make the appearance argument. So far, you've failed to provide much in the way of evidence of real world editing failing in ACR while being demonstrably better in some other app. And different is not better, it's just different. Go any images where ACR fails and some other app shines?

Hi Jeff,

I've attached some image (!) samples for you, that show the difference in handling of Contrast changes. Also note that it's not only the different look (which of course can be further processed to look more similar), but it is also about the workflow which IMHO allows to better achieve a certain creative intent, because we do not need to compensate color saturation issues caused by Contrast adjustments as much. We can simply adjust Contrast if we want to adjust Contrast, and adjust Saturation if we want to adjust Saturation, and even the order in which we do that is not important.

It is difficult to create an exact one-on-one comparison between different applications, due to differences in camera profiles and color engines. So the images will already look different, but I've tried to make the baseline images look somewhat similar.

First the before and after images of Capture One Pro version 9. When I can find some more time, I'll add Lightroom or ACR results in a followup post.

I've sampled one spot in the sky, and noted the RGB, Lab, and HSB coordinates of the resulting JPEGs, first before contrast boost and then after a contrast boost of +50. That's the maximum of the scale, which probably is a different scale than ACR/LR anyway.

** Blue Sky **
RGB = [139, 165, 192] --> RGB = [161, 194, 229]
Lab = [66, -5, -17] --> Lab = [77, -5, -21]
HSB = [211°, 28%, 75%] --> HSB = [211°, 30%, 90%]

The contrast curve (probably) has a different shape in Capture One, so let's not focus on the absolute values, but the relative change. Easiest for most people is maybe to compare the HSB (Hue, Saturation, Brightness) coordinates. Hue is not changed after the Contrast boost. We can see a significant increase of the Brightness, as expected for a boost in contrast of brighter colors in a larger scene. Saturation only scores a modest increase of 2%, despite the Brightness boost by 15%.

Of course highlight detail gets severely compressed by such an extreme contrast boost, so one would usually correct that with the other controls but that's not what we're investigating in this thread (nor would I use such an extreme contrast boost in normal practice). I'll repeat, the extreme adjustments are done to see if and how the algorithms fall apart, because small changes will even look acceptable with a crappy algorithm and that would not tell us much.

The other crops from the same scene are of two different grass tones, one dark shaded tone, and one transilluminated more yellow tone.

** Grass darker **
RGB = [64, 83, 1] --> RGB = [36, 52, 3]
Lab = [33, -16, 39] --> Lab = [19, -13, 26]
HSB = [74°, 99%, 33%] --> HSB = [80°, 94%, 20%]

** Grass lighter **
RGB = [144, 174, 1] --> RGB = [174, 217, 6]
Lab = [67, -25, 67] --> Lab = [81, -31, 78]
HSB = [70°, 99%, 68%] --> HSB = [72°, 97%, 85%]

Here we see a change in Hue, more for the darker grass than for the lighter grass. The darker Green gets darker, and the lighter Green gets lighter, as expected for a contrast boost. We see that both Greens get Desaturated (!), the dark patch more (to avoid oversaturation due to the already low blue channel contribution).

When I have a bit more time I'll do a similar run with Lightroom, but people can of course already have a look at how their own images handle extreme contrast boosts of sky blue and grass/foliage.

Cheers,
Bart

[Tim Lookingbill]

Bart, your image samples don't prove anything. Good grief. Why would someone apply +50 contrast to a cloud image whose overall luminance is above middle gray. And no one would just do the same with the greenery image.

I take it you want the simplest turnkey twiddle knob to do the same across all images without any sense of what the real scene looked like as a starting point. Those samples are unrealistically predetermined to fail adding a contrast boost.

Go back and try again.

[Bart_van_der_Wolf]

Bart, your image samples don't prove anything. Good grief. Why would someone apply +50 contrast to a cloud image whose overall luminance is above middle gray. And no one would just do the same with the greenery image.

Tim,

Reading comprehension appears to be harder for some than I thought. I already explained, several times, that small changes will lead to small issues, so IF ONE WANTS TO LEARN something about the robustness of the algorithms used, exaggeration helps. If extreme corrections do not lead to the kind of issues that this thread is about, then lesser adjustments will do even better.

Cheers,
Bart

[digitaldog]

Bart, your image samples don't prove anything.

That's pretty harsh Tim. I don't agree  . Bart clearly demonstrated that a raw file will appear different depending on the raw processor. Maybe that's a big idea/concept for some, for the rest of us, not at all. Otherwise, I don't think he proved anything (so we're then in violent agreement!).

[Tim Lookingbill]

Tim,

Reading comprehension appears to be harder for some than I thought. I already explained, several times, that small changes will lead to small issues, so IF ONE WANTS TO LEARN something about the robustness of the algorithms used, exaggeration helps. If extreme corrections do not lead to the kind of issues that this thread is about, then lesser adjustments will do even better.

Cheers,
Bart

Not reading comprehension, Bart. Just lack of patience and time to read that long complicated outline of yours. So I had to assume going by the title of this thread that you were demonstrating contrast induced saturation. I still wouldn't know what you said even if I did read all of it because it's just too damn long and complicated to focus on. I read your other long posts and I still kept reading the same obvious stuff I already know. I just assumed you were pointing out more of the obvious.

The time it would take to read all that and understand it, I just improved the look of your grass/concrete step image without any saturation issues in ACR 6.7.

I still don't see with your points being made how you're making editing images any easier and less complicated. I still make better looking images without all that analysis. Why can't you?

[Tim Lookingbill]

We see that both Greens get Desaturated (!), the dark patch more (to avoid oversaturation due to the already low blue channel contribution).

Well if one of us has reading comprehension issues then the other has color perception issues. I'm not seeing that green get desaturated, Bart. I see more saturated green compared to the original and since you didn't white balance the image (concrete sidewalks aren't greenish yellow in bright sunlight) you've actually compounded the appearance of over saturated greens with the added contrast. 

White balance plays hand in hand with saturation appearance and the amount of its increase due to increased contrast.

Could you post an image that is finished and looks good to you because you edited with software that doesn't increase saturation when adding contrast? Post a before and after. I would once like to see an image that looks good with this technique.

[Mark D Segal]

Could you post an image that is finished and looks good to you because you edited with software that doesn't increase saturation when adding contrast? Post a before and after. I would once like to see an image that looks good with this technique.

Good question! Without being blasphemous, it could be a Frosty Friday in Hell before one like that occurs, but one must always ask just in case.........."never say never" :-)

[Bart_van_der_Wolf]

Well if one of us has reading comprehension issues then the other has color perception issues. I'm not seeing that green get desaturated, Bart. I see more saturated green compared to the original ...

Tim, when the S  (is Saturation) in HSB coordinates goes from 99% to 94% for darker grass (HSB = [74°, 99%, 33%] --> HSB = [80°, 94%, 20%]), and for lighter grass from 99% to 97% (HSB = [70°, 99%, 68%] --> HSB = [72°, 97%, 85%]), then saturation has decreased.

What you have just demonstrated is that we cannot rely on our eyes to estimate absolute colors/values, and even in comparison we can miss the mark. What did change was the Contrast and thus the Brightness depending where on the various brightnesses land on the contrast curve. When we change the Brightness of a Color, even if we would keep the Saturation the same, it seems to have changed. That's how human vision works. The last thing we need when changing Contrast, is another Saturation boost, which we would then have to reduce by another editing effort (which is quite doable as you've shown in your earlier example, but requires adding another edit step).

This thread is not about if we can correct Saturation boosts that result from Contrast changes, but why should we need to do that if the software could largely avoid the excessive Saturation boost (or reduction) to begin with.

... and since you didn't white balance the image (concrete sidewalks aren't greenish yellow in bright sunlight) you've actually compounded the appearance of over saturated greens with the added contrast.

Wrong again. The concrete was not pure grey*, and the White-balancing was exactly the same as for the white of the clouds. Had I color-balanced on the concrete, then the clouds would have been too blue. I've attached an HDR rendering of the scene that I've posted for another purpose in another thread before.
*) Exposure brackets for that HDRI were used for the crops I posted here and they are of one and the same scene, and Art object (a WW II bunker, number 599, from 1940 that was sawed in half).

White balance goes hand in hand with the appearance of Saturation and the amount of its perceived increase due to increased contrast.

Could you post an image that is finished and looks good to you because you edited with software that doesn't increase saturation when adding contrast? Post a before and after. I would once like to see an image that looks good with this technique.

For the purpose of demonstration I'll humor you. Attached are a before, and an after +25 Contrast adjustment image, made in Capture One Pro version 9.2.0. Also, it is not a technique, but a single (more intelligent) Contrast slider control.

Again, there is generally only a very modest change in Saturation, and virtually no color Hue shift, compared to the significant change in contrast (higher for brights, lower for darks). And because our eyes can fool us, I've again added some color sample readouts, taken from the images before final conversion, in this case from AdobeRGB to sRGB, for display.

* Sky *
RGB = [132, 141, 175] --> [144, 154, 193]
Lab = [59, 1, -20] --> [64, 1, -23]
HSB = [227°, 24%, 69%] --> [227°, 25%, 76%]

* Grass dark *
RGB = [61, 62, 38] --> [47, 47, 28]
Lab = [24, -4, 17] --> [17, -4, 15]
HSB = [61°, 38%, 24%] --> [61°, 41%, 19%]

* Building *
RGB = [158, 118, 100] --> [174, 126, 105]
Lab = [55, 20, 20] --> [59, 24, 23]
HSB = [19°, 37%, 62%] --> [18°, 40%, 68%]

Now you'll probably say that the contrast has become too high, or find an other excuse to reject the image quality (like the sky being a tad too magenta, which might be true but that was the same for both images), but I did not have to adjust the Saturation like you had to do in the example that you posted earlier. I could have, and would only have needed a very small amount to achieve that, without again changing the contrast as a result, causing to revisit that, which in turn .....

The question that this thread is about is simple, when we choose to change the contrast, then why do we also need to adjust the Saturation back to where it already was???

Cheers,
Bart

[digitaldog]

The question that this thread is about is simple, when we choose to change the contrast, then why do we also need to adjust the Saturation back to where it already was???

The answer is equally simple. Because most of the time, it LOOKS better. And if it doesn't, there's a fix (adjustment) for that.

[Bart_van_der_Wolf]

The answer is equally simple. Because most of the time, it LOOKS better. And if it doesn't, there's a fix (adjustment) for that.

Hi Andrew,

It sounds like you like to dance. Two steps forward, one step back, to get where you wanted to be in the first place.
Seems like a waste of time to me, unless you like to dance instead of saving the time for other things (aka productivity and higher output quality).

Cheers,
Bart

[digitaldog]

It sounds like you like to dance. Two steps forward, one step back, to get where you wanted to be in the first place.

Incorrect on both points. Again, most, if not all the time for myself and perhaps tens of thousands of Adobe users, the default behavior is what looks better! Was Jeff's description of what Thomas found and thus coded unclear to you? Seems so. Like any default position, there are times it's not appropriate and IF so, there's an easy fix. Meanwhile, the behavior you propose, that doesn't LOOK as good as what you are arguing against, would require more dancing, steps, more of the time.

[Bart_van_der_Wolf]

Incorrect on both points.

Maybe you should try it then. Dancing can be good exercise and it can be fun (depending on the company one keeps). But it leads you nowhere, essentially just going round in circles. In my younger years I received a gold star recognition for my Ballroom dancing skills.

Again, most, if not all the time for myself and perhaps tens of thousands of Adobe users, the default behavior is what looks better!

Well, in that case I'd hurry and convince Eric Chan, before he improves the Raw conversion engine...

Cheers,
Bart

[digitaldog]

Maybe you should try it then. Dancing can be good exercise and it can be fun (depending on the company one keeps).

As someone who walks/runs over 5 miles a day, I'm good. Just as I and many, many others are more than good with the current Adobe behavior that again, was produced BY design because it LOOKS better 9+ times out of ten. And when it doesn't, easy to adjust.

Well, in that case I'd hurry and convince Eric Chan, before he improves the Raw conversion engine...

I hope he and Thomas continue to do so and expect they both will. Meanwhile, you're not getting much traction around these parts with your concepts of Contrast and Saturation but hey, it's your time to waste.

[Peter_DL]

Recommended reading: Tone curves and camera profiles
by Anders Torger

quote: >>So then all is good with the tone curves applied by typical raw converters? No. In fact if we’re into a neutral and realistic starting point it’s not good at all. Most converters apply a pure RGB curve which has little to do with perceptual accuracy. Lightroom and many DNG raw converters apply a slightly different RGB curve that reduces hue shift problems (HSV hue is kept constant), but it’s still in most situations almost identical in look to a pure RGB curve. It varies between converters in which RGB space this curve is applied, which also affects the result. In Lightroom/DNG it’s always applied in the huge linear ProPhoto color space, while in many ICC raw converters it’s applied in a smaller color space.

Let’s start with the RGB tone curve problems. It will increase saturation more than is reasonable to compensate for Stevens and Hunt effects, so you get a saturated look. You might like that, but it’s not realistic. Another problem is that for highly saturated colors one or more channels may reach into the compressed sections in highlights or shadows and that leads to a non-linear change of color, that is you get a hue shift. Typically the desired lightening and desaturation effect (transition into clipping) masks the hue shift so it’s not a huge problem, but it’s there.

Then there is the color space problem. If the RGB tone curve is applied in a large color space such as one with ProPhoto primaries (like in the DNG case) one or more channels can be pushed outside the output color space (typically sRGB or AdobeRGB) so we get clipping and thus a quite large hue shift. Some raw converters partially repair this through gamut mapping (Lightroom does), but still there may be a residual hue shift.

To battle the various RGB tone curve issues bundled profiles typically have various subjective adjustments to counter curve issues. For example the profile may desaturate high saturation reds to avoid color space clipping. Naturally this means that the same profile used with a linear curve will produce too little saturation in the reds. That is a profile must be specifically designed for the intended curve.

I think this is bad design. In fact one could argue that staying with RGB curves (and similar) has inhibited the development of good profiling tools and makes it unnecessarily hard to get natural colors in our photos.

It doesn’t have to be this way, the RGB tone curve is legacy from the 1990s when its low computational cost was one of the reasons to use it. It can also be seen as a nostalgic connection to film photography. In the film days the film had to produce the subjective look too, so exaggerated contrast and saturation were desirable properties. This thinking has been kept in most raw converters today despite that we have all possibilities to start from a neutral look and design our own on top rather than relying on bundled looks. The RGB tone curve produces a saturated look that many like to have in their end result, but as said it still doesn’t work well for profiles that aren’t specifically adapted for it.

Using a DCamProf neutral linear profile and applying and RGB tone curve will produce a garish look. As we will see, the solution to this problem is to use DCamProf’s built-in neutral tone reproduction operator.<<.

[brandon]

Meanwhile, you're not getting much traction around these parts with your concepts of Contrast and Saturation but hey, it's your time to waste.

Actually its a really interesting post, for the views and understanding of whats "under the hood" and what other image characteristics are affected by alterations in contrast (which I think most of us manipulate on most images) and how that differs with different raw developers (and between their process versions). Of course preference is subjective (thankfully), but the suggestion that what is best is what was first done (eg by ACR, lightroom and C1 up to version , or because most people are happy with what they are used to (a proxy being user numbers) seems a bit stuck in the mud. Thanks Bart for as always providing detailed explanation (often over my head, but that's my problem), and examples to illustrate those discussions. Please dont consider doing so a waste of time as suggested.
Cheers
Brandon
PS less interesting is reading how "short" some are in their comments (this thread and others)

[Schewe]

Recommended reading: Tone curves and camera profiles
by Anders Torger

Not saying that Anders has an agenda but he clearly doesn't like the default tone curve in DNG Profiles (something I'm surprised that Bart isn't jumping up and down and yelling about) but I'm not sure that recommended reading is 100% on point. Anders wanted to create a DNG profile that didn't apply a tone curve by default. There are ways around that by using DNG Profile Editor and defeating the the builtin tone curve.

Course then you have the problem of either living with ACR's toning controls and understand there are a LOT of tone controls above and beyond Contrast-which truth be told I rarely if ever use or process the raw image with a linear tone curve and manipulate further in Photoshop or your other non-Adobe manipulator of choice.

But I think it would be better to stick to the "ACR's Contrast control suck" argument.