Mark Segal has written an excellent essay on this subject and discusses certain controversial claims made by a prominent author on this subject.
The following quote sums up much of the findings:
"Thomas Knoll was asked how these Curves algorithms affect saturation and hue. His reply indicated that in both Photoshop and CR the Curves are designed to moderately boost saturation with contrast, resulting in saturation effects that resemble those from film tone curves, which many years of looking at photographs from film has taught our visual perception to like. He said the programming is easier without the saturation boost, but included it because after extensive testing they determined the results are generally not pleasing without it."
Hmm. I read through that and was generally no wiser at the end. I must read it again later. I had been using a Point Curve default which is halfway between Linear and Medium, and have now moved back to the default of Medium. Colours do seem to be slightly richer now, which is surprising. I think I'll also experiment with higher Blacks values for a while, as that tends to help (presumably by increasing contrast).
The reason I ask is because I have a friend with an identical camera who uses Lightroom, and he seems to get richer colours. I've actually gone to the effort of creating a custom profile with a ColorChecker and DNGPE (based on Adobe Standard), but I still seem to be pushing up the Vibrance and Saturation on a lot of photos, where his settings seem to be default, and have higher Blacks.
Just a cautionary warning: The link to the essay in post #1 is a 2MB PDF that starts to download as soon as you click on the link. I hate it when that happens without warning. :/
I don't know if this is correct but it's how I think of it.
In space you need three points to define a specific point. If you think of a color gamut as being a "space" then you need three points to define a point in that space. But you can use more than one coordinate system to define point in this space. For example one could be RGB and another would be HSL. When working in ACR (or most of the time in PS) you are working with the RGB coordinate system to adjust (map) points from one place to another. If you are adjusting points in RGB space to affect tone/luminosity you will also be affecting the other dimensions is that space--hue and saturation.
Like I said, I'm not sure this stands up to scrutiny; but, it's how I think of it.
Basically if you increase contrast, you increase saturation. The inverse is also true, i.e., if you decrease contrast, you decrease saturation.
Aha! So, a shallower custom tone curve *would* decrease saturation. That explains a lot.
I was reluctant to set a non-zero default for saturation, because I was convinced there was a good reason for my colours being a bit weak.
Thanks for putting it so simply, Eric.
Why this is considered pleasing goes even further back to the photorealist painters of old like the Dutch masters to create the illusion of 3D renderings on a 2D flat canvas same as photography and prints. Using a luminance curve instead of a regular RGB composite curve to increase contrast is like the master painters mixing primarily black neutral paint to increase density which kills depth and richness in mids and shadows and was a big "NO, NO" back then as it is today.
You can do a test to illustrate this effect more clearly by creating two duplicate sets of files containing identical color swatches of your choosing. Make a selection of each half of each swatch in each file with one file's selection placed on its own blend layer set to Luminosity and apply the same darkening curve or use Level's gamma adjust and see how hue/saturation is affected differently between the two. Then decide which one has more depth and will pop off the page in an inkjet print. The regular RGB version will most likely be the one.
Exposure, white balance and lens coating and quality of glass have a big affect on locking in the correct and balanced combination of density with evenly distributed levels of saturation so the user can apply their own tone map to get the results they want from the start without getting saturation hot spots which can ruin the 3D effect. Calibrating using a CC chart helps insure this but isn't perfect all the time.
The PDF seems to be removed. Anyone knows an alternative source, please?
I've contacted Mark, who contacted Michael, who restored the files yesterday.
>> Basically if you increase contrast, you increase saturation. The inverse is also true, i.e., if you decrease contrast, you decrease saturation.
How about an orthogonal definition: HS-L* or HS-Y (see Simon Tindemans).
Regarding Mark's article, there also was a broad discussion:
Yes, those are possible to implement, but not always desirable. As is often the case, you will find they work better on some images, and worse on others.