Separating saturation from a contrast adjustment

[DanK]

I don't know if folks will find this interesting, but here is an example of something that came up recently in another thread. If you apply a simple curves adjustment in the software with which I am familiar, you increase contrast in the three color channels and therefore increase saturation as well as contrast. Sometimes you might want that, but sometimes you might want to keep the two adjustments separate. I haven't examined other related adjustments, such as a levels adjustment, but I think it would behave the same.

First, the original, washed out, foggy image:



Now, the fairly extreme curves adjustment:



Now, the effect of applying this curves adjustment with a normal blending mode (photoshop):



And last, the same curves adjustment set to luminosity blend mode:



The point is not whether one of these is superior. This is just an illustration of the difference between applying a contrast adjustment in the normal mode and applying it only to luminosity. I believe that you would get the same thing as the last image if you switched to LAB mode and set the curves adjustment to the "lightness" channel, but I didn't test that.

[Stagecoach]

Dan, I have also experimented (using similar images) with using a luminosity layer to reduce the increased saturation when using normal layer mode.

What I'm not sure of is whether it is or why it is 'advantageous' to do this rather than simply reduce saturation if not wanted.

[chauncey]

FWIW...using curves/levels is but one way to accomplish your goal. The hue saturation adjustment
layer will allow more wide ranging latitude...using the B&W adjustment layer set to luminosity allows
the same thing, sans the boost in saturation.

[DanK]

I think it can provide more control to separate them.

[ajohnw]

I'm inclined to agree Dan but as far as i know I don't generally use anything but luminosity on either curves or levels.

There is another way expanding on WM's post. Duplicate layer and desaturate, overlay mode. Duplicate that and invert, overlay mode. One brightens dark parts and the other darkens light parts so effectively the brightness range has been flattened, each been set by the opacity sliders. In the case of mist it then needs some form of contrast increase layer. This is using a new from visible of the above in softlight mode which of course will increase saturation as well. The mist tends to loose out due to the brightness range reduction. Something to experiment with as it seems to be very good at clearing mist.



I can do this in a Linux package very quickly but it's not a good idea to use it on an image that has had any sharpening at all. Different technique for adding the contrast.

John
-

[Ashwin]

I think, if you want to increase or decrease t he saturation only. You should select Image > adjustment > hue/saturation and in the resultant window that opens, move the slider saturation as you desire[IMG][/IMG]

[IzzieK]

As with William, I also use Hue and Saturation so this may be another method. Another one is the use of Luminosity Masks.

[ajohnw]

The problem with the layers approach I mentioned rather than the one WM mentions is that 2 such layers in theory should cancel each other out but it does give a choice of where the brightness range is flattened which is the part that tends to get rid of mist if it's applied aggressively and the overall contrast is subsequently boosted some how. i'm not to happy with the one I mentioned but wonder what happens if other things are done to the contrast layer or maybe there is a method that doesn't change saturation much.

Interesting post for me. I suspect this area is why many PS shots look as they tend to look to me as curves is messing with saturation as well and not giving a natural final look. There are some noticeable exceptions - Christina for one springs to mind and some others. People who use OS software don't tend to go the same way but it's easy to add something similar with a softlight later in the GIMP.

John
-

[John 2]

The problem with the layers approach I mentioned rather than the one WM mentions is that 2 such layers in theory should cancel each other out but it does give a choice of where the brightness range is flattened which is the part that tends to get rid of mist if it's applied aggressively and the overall contrast is subsequently boosted some how. i'm not to happy with the one I mentioned but wonder what happens if other things are done to the contrast layer or maybe there is a method that doesn't change saturation much.

Interesting post for me. I suspect this area is why many PS shots look as they tend to look to me as curves is messing with saturation as well and not giving a natural final look. There are some noticeable exceptions - Christina for one springs to mind and some others. People who use OS software don't tend to go the same way but it's easy to add something similar with a softlight later in the GIMP.

John
-

Interesting and useful. I have to admit that there are occasions in which after processing an image in the standard way in PS, I look at the result and it just doesn't look right. No amount of tweaking thereafter helps, it just seems out of balance tonally. It has never occurred to me that separating contrast from saturation may provide more control in avoiding the problem. Some experimentation seems to be on the cards.

@John - presumably, additional control is available using your method by altering the opacity of the individual layers?

[chauncey]

Oftentimes, when I've amassed a lot of layers, I will turn them off to gauge the effect on the final image.

[DanK]

Ashwin and Izzie,

I must not have been clear. The point of my post was not that one can use curves to adjust saturation. The point was that if you simply apply a curves adjustment, you will be adjusting saturation whether you want to be or not. The point of using a luminosity blend mode (which acts solely on luminance, not on colors) is that it allows you to adjust contrast without adjusting saturation.

I first ran into this issue some time ago, while editing this image of a parrot tulip:



The reds are not clipped, but I found that after editing, I lost a lot of the detail in the reds. There were a number of problems, but the biggest one was that I was applying a contrast adjustment via the curves tool, without switching to luminosity blend mode. The increase in saturation was obscuring the detail in the reds.

I just checked, and the same happens with a levels adjustment, but the effect is far slighter.

I have just started playing around with LAB color and this seems to take care of the problem as well, because in the LAB color space, you have to pick which channel the curves tool acts on, and you can pick "lightness," which is essentially luminance.

Dan

[ajohnw]

Interesting and useful. I have to admit that there are occasions in which after processing an image in the standard way in PS, I look at the result and it just doesn't look right. No amount of tweaking thereafter helps, it just seems out of balance tonally. It has never occurred to me that separating contrast from saturation may provide more control in avoiding the problem. Some experimentation seems to be on the cards.

@John - presumably, additional control is available using your method by altering the opacity of the individual layers?

It does give control but not to the same extent as a Linux package I use which has a slider called brightness distribution - flatten and it does just do that, The histogram literally flattens. It also tends to extend. This for instance is 75% of it, very heavy and the sort of level that seems to clear mist. Any more would introduce black clipping. Might be a bit anyway.



To me now the image looks clearly sharpened which I find is not a good idea if contrast is then applied some how. ideally leaving saturation alone. I have no idea how the package does this so keep trying to do it with layers. I'd be glad of anyone's thoughts.

So people can compare this is Dan's original misty image.



John
-

[ajohnw]

Should add the next thing I will try is contrast adjustments on an overlaid desaturated copy either inverted or not inverted.

John
-

[rpcrowe]

I used NIK Viveza:
Structure +50
Contrast +10
Shadow Adjust -10

[DanK]

I used NIK Viveza:
Structure +50
Contrast +10
Shadow Adjust -10

I think a comparison with my first three suggests that the Viveza contrast adjustment behaves like a regular (not luminosity) curve adjustment: it increased saturation. A purer test would be to apply a substantial contrast adjustment with Viveza without any other adjustments. I just took a lunch break, so I did that. here is the image with nothing but a +50 Viveza contrast boost. Clearly, it increases saturation:

[ajohnw]

I had never thought of doing a web search for the technique I used. For PS it is described here. Simple brightness and contrast changes work as well.

http://www.luminous-landscape.com/tu..._masking.shtml

Actually small amounts can be useful on many shots. Here's a sub 1min don't care too much use of it on the GIMP. I even sharpened it a bit after adjustment, maybe a touch too much. I just used brightness and contrast on it. Curves would obviously allow a fair amount of variation.



John
-

[ajohnw]

Being fair - slight s applied to the above.



John
-

[IzzieK]

John and Richard, these are interesting image to experiment. I forgot about the LAB PP, Dan. Sorry...There are just so many options including Nik and I suppose On One has one to which one can come up with when it come to saturation and contrast and all that jazz, eh?...

[xpatUSA]

Me being me, I just had to put some figures to this subject, in other words I was interested in bow much the saturation increases. So I took a shot of the top row of the Macbeth card and listed the RawTherapee HSV values after a neutral conversion to 16-bit bit TIFF, ProPhoto working space. (The top row has more natural colors than the others and is somewhat less saturated to start with).

Then I applied a luminosity tone curve which was the same shape as that posted in the OP. Sure enough, the saturation of each patch increased significantly. I measured the fractional increase for each patch and then found the mean of those 6 fractions.

The result was that the average fractional increase in saturation was 1.336 (I'm avoiding the term "percent" because those are the units of saturation per se). Standard deviation was 0.14. Do pardon the statistics but it's a lot easy than typing in all my measurements.

[DanK]

Hi Ted,

Interesting. Glad to see that I understood the L curve in LAB space correctly. No need to apologize for the statistics, at least to me.

One question. You wrote this:

Then I applied a luminosity tone curve which was the same shape as that posted in the OP.

I don't know Raw therapee, so I don't know what you mean by a 'luminosity tone curve.' In photoshop, if you apply a tone curve without doing anything else, it applies the change to the color channels. You only get a luminosity-only curve by setting the blend mode to luminosity, rather than RGB, or by switching to LAB as you did. I'm assuming that the luminosity curve you are referring to is the garden-variety adjustment that acts on the color channels. Is that right?

Dan