Shadow/Highlights tech question

[jfriend]

I'm a frequent user of the shadow/highlights feature in CS2 and there are many outdoor photos with so much dynamic range that they are really helped by an appropriate amount of shadow/highlights.

But, I'm trying to understand exactly what the different settings in the dialog do. I know sort of what they do, but am interested in more technical detail.

To refresh you memory, here's the dialog:


The core adjustments are amount, width and radius. At the 50,000 foot level, width controls how much of the tonal region is affected, radius is something related to a blend distance and amount is how much of a correction is applied. But I was hoping for a little more technical definition of these.

Here are some possible ways it works. I'm interested if anyone knows which of these might be right or close or if there's anything written on the web for how it really works. Imagine a scenario where I'm doing a shadow adjustment with width=50% and radius = 30. Anyone know which of these best describes how it works:

• Only tones from 0-128 are affected (width = 50%) and the change to the affected pixels are blended in at a radius of 30 pixels from the change. Pixels at the darker end are affected more than mid-tone pixels with the effect fading to nothing at tonal value 128.
• Tones from 0-128 are equally affected with the effect fading to zero in the range of 128-something higher than 128 (perhaps guided by the radius). The effect is then blended using the radius of 30 pixels.
• Tones from 0-128 are equally affected and no tones above 128 are affected except if impacted by the blend using the radius of 30 pixels.

The kinds of questions I'd like to be able to answer are:

• If I want to isolate the effect to only tones 0-80 and stay out of the mid-range tones (except for the impact of blending), what would I set the width to?
• How do you figure out what's a good radius to set? When should you be using a small radius and when a large radius?
• When you have a light/dark boundary (dark foreground trees against a medium tone sky, for example), what settings are going to help prevent the visible halo you will some get along the sky border when you try to apply a shadow adjustment to the trees.

I'm also curious about some of the other settings in this dialog:

• What does "color correction" do? Since I pretty muchy always apply S/H on a dupped layer, is it desirable to just change the blend mode of the S/H layer to luminosity and forget about the color correction setting?
• What does "midtone contrast" do? Is it any different than applying your own curve after the fact?
• What do black clip and white clip do?

Any help in further understanding this is much appreciated. Thanks.

[DavidTO]

Originally Posted by jfriend

What does "color correction" do? Since I pretty muchy always apply S/H on a dupped layer, is it desirable to just change the blend mode of the S/H layer to luminosity and forget about the color correction setting?

This one I think I know. By definition, the closer you get to black or white in RGB, the less color there is in it. Black is R0G0B0, completely neutral, so if you raise a near-black shadow it will get lighter, but will be desaturated. Color correction gives you an opportunity to compensate for this.

(How'd I do? )

[jfriend]

Originally Posted by DavidTO

This one I think I know. By definition, the closer you get to black or white in RGB, the less color there is in it. Black is R0G0B0, completely neutral, so if you raise a near-black shadow it will get lighter, but will be desaturated. Color correction gives you an opportunity to compensate for this.

Thanks for helping. More questions...

I get that the brights and darks are neutral in RGB (that's one of the unique things about LAB that you can have bright or dark with color), but if something is essentially neutral in a deep shadow in RGB, how would it know what color to make it when it makes it brighter?

[pathfinder]

This aspect of RGB, is why the shadow/highlight tool works so much better on the L channel in LAB. This should help avoid the color changes incurred in RGB, I believe.

That is they only way I use shadow/highlight.

And I prefer to use ONLY shadow or Highlight if possible, rather than both. I restrict shadow to 15% or less and highlight to 5% or less if I am using both.

[jfriend]

Originally Posted by pathfinder

This aspect of RGB, is why the shadow/highlight tool works so much better on the L channel in LAB. This should help avoid the color changes incurred in RGB, I believe.

That is they only way I use shadow/highlight.

And I prefer to use ONLY shadow or Highlight if possible, rather than both. I restrict shadow to 15% or less and highlight to 5% or less if I am using both.

If I'm taking the document to LAB for other reasons, I'll definitely use S/H on the L channel (just like I sharpen the L channel). But, if I'm not going to LAB for other reasons, I will often just do S/H on a dupped layer set to a luminosity blend mode (it seems to accomplish mostly the same thing) and then I don't have to switch modes just for this operation.

[jfriend]

For those that are interested, I got a couple good references in another online forum and pretty much figured out the answers to my questions. This article describes, in the words of a Photoshop engineer, how the width setting works. And, this PDF document describes how you can set up a test image to be able to run your own experiments to see exactly how the shadow/highlight feature works. From the first article, I came to these conclusions:

Tones in your image are affected by the shadow setting in a linear progression with max effect at the shadow point and no effect at the tonal value determined by the width percentage. So if you set it to 50%, you get max effect at a tone of 1 and no effect at a tone of 128 with a linear ramp between the two. You'd get half the effect at a tonal value of 64, one quarter the effect at a tonal value of 32 and so on.

So, the width setting is the maximum amount of tone that you want affected by the setting with the effect ramping linearly down to end at that value. It has no effect at tones beyond the width value.

So, if you want to protect mid-tones from any change, you could use something like a width setting of 30%. This is, of course, all subject to the blending radius so lots of other tones will still be affected within the the blend radius of a changed area.

The color correction slider is used to increase or reduce the saturation in the areas affected by the change. If you use shadow/highlights on the L channel or on a luminosity blend mode or fade to luminosity, you presumably don't need this control.

I still don't quite understand how the clipping controls work, but it sounds like it's best to ignore them in S/H and use the traditional curves or levels before you do the S/H adjustment to set your white point and black point independently.

The mid-tone contrast adjustment is similar to a curve adjustment on the mid-tones. Since it's only a slider, it's less flexible than a real curve - I don't see any reason to use this, more priimitive, slider.

[Andy]

I too, go very lightly on this tool, but it has definite uses, and should not be ignored. 15-18% max on the shadow tool, for me.

If working on Shadows, decreasing tonal width works on only the darkest spots.. increasing it brings up a wider range of shadows. Go fullout, and it works on the midtone area shadows.

Radius amount affects how "wide" in pixels the adjustment is. I find that some images can take a lot more highlight adjustment than others.


The main thing, is to ALWAYS do this adjustment on a new layer, that way you can easily mask or reduce the opacity.


DavidTO's Shadow Recovery Tutorial

Rutt's Highlight Recovery Tutotorial

[jfriend]

Originally Posted by Andy

I too, go very lightly on this tool, but it has definite uses, and should not be ignored. 15-18% max on the shadow tool, for me.

If working on Shadows, decreasing tonal width works on only the darkest spots.. increasing it brings up a wider range of shadows. Go fullout, and it works on the midtone area shadows.

Radius amount affects how "wide" in pixels the adjustment is. I find that some images can take a lot more highlight adjustment than others. The main thing, is to ALWAYS do this adjustment on a new layer, that way you can easily mask or reduce the opacity.


DavidTO's Shadow Recovery Tutorial

Rutt's Highlight Recovery Tutotorial

I agree that it is easy to overdo the shadow adjustment in this tool and it creates a funny and identifiable look when it's overdone. How much you can go seems very image-dependent to me and it also seems to depend a lot on both the radius and the width you are operating at. I've found some images that I could turn both the radius and width down and they could benefit from quite a bit of shadow amount (30-40%), probably because I was only affecting a narrow tonal range with a smaller radius.

It's also worth saying that this tool should not be a substitute for curves. I try to do my curve adjustments first (sometimes raising shadows on the curve) and see how that works on your image. Then, only when I don't get the result I want with curves do I try adding in a shadow/highlight adjustment.

On the highlight side of things, it's sometimes real magic how it can take highlights that are not actually blown, but are so bright that you can't see any detail in them and bring detail right back. I found this particularly true in pictures of my kids in the snow (the detail that's coming back is in the snow).

[rutt]

Another pearl of wisdom from Dan Margulis' LAB book. Recover highlights in RGB and shadows in LAB. The contrast range is mapped differently in the two color spaces and LAB devotes more tonal range to shadows whereas RGB devotes more to highlights. So the shadow recovery tool has more space to work in LAB and the highlight recovery tool has more space to work in RGB. Since I learned this, I've found I can get away with larger amounts and tonal ranges and get better results. PF is also right that using the shadow tool on L channel prevents to possibility of any color shift.

If you want the same effect for highlights, build a luminosity blending layer (as in the portrait technique), apply the highlight recovery tool to it in RGB, and then convert to LAB before flattening.

[Andy]

Originally Posted by jfriend

On the highlight side of things, it's sometimes real magic how it can take highlights that are not actually blown, but are so bright that you can't see any detail in them and bring detail right back. I found this particularly true in pictures of my kids in the snow (the detail that's coming back is in the snow).

and esp great when you don't have a RAW file to recover from. Look here

http://www.dgrin.com/showthread.php?t=25720 posts 2 and 3...
pretty darn good recovery of an otherwise totally blown bird.

[rutt]

And, I've also learned to recover in a duplicate layer so I can use blending options or a layer mask or just opacity to fine tune afterwards.

[jfriend]

Originally Posted by rutt

Another pearl of wisdom from Dan Margulis' LAB book. Recover highlights in RGB and shadows in LAB. The contrast range is mapped differently in the two color spaces and LAB devotes more tonal range to shadows whereas RGB devotes more to highlights. So the shadow recovery tool has more space to work in LAB and the highlight recovery tool has more space to work in RGB. Since I learned this, I've found I can get away with larger amounts and tonal ranges and get better results. PF is also right that using the shadow tool on L channel prevents to possibility of any color shift.

If you want the same effect for highlights, build a luminosity blending layer (as in the portrait technique), apply the highlight recovery tool to it in RGB, and then convert to LAB before flattening.

Thanks for pointing this out Rutt.

I knew that it was better to do operations like S/H on the L channel, but until I just found and reread pages 146-148 in Dan's book, I didn't really know why. I've read those pages before, but it didn't stick until now.

To summarize it for others reading this, the (127,127,127) RGB midpoint tone is significantly above mid-range in the L channel (similar to 157,157,157 according to a quick experiment in CS2). What that means is that the L channel has a lot more range to deal with shadows than RGB does (more tonal values devoted to shadows) thus they are spaced out further and easier to deal with.

The inverse is true for highlights. The highlights are compressed in the L channel so they are easier to deal with in RGB.

I don't know why LAB was originally designed this way, but it makes a lot of sense for our eyes because we can see shadow variations a lot more than we can see highlight variations.

[Andy]

This is a really good discussion thread - and so I've linked back to here from the two tutorials on http://dgrin.smugmug.com/How-To/143180

Thanks Guys!

[dandill]

Originally Posted by Andy

This is a really good discussion thread - and so I've linked back to here from the two tutorials on http://dgrin.smugmug.com/How-To/143180

This tutorials portal is a wonderful resource.

A suggestion on the tutorials: As new ones appear or old ones are updated, could the date of their last update be added somewhere?

[Andy]

Originally Posted by dandill

This tutorials portal is a wonderful resource.

A suggestion on the tutorials: As new ones appear or old ones are updated, could the date of their last update be added somewhere?

I've toyed with leaving the updated date on the descriptions. But I like the cleaner look..

Have you tried subscribing our RSS Feed?

[jfriend]

Originally Posted by jfriend

I'm a frequent user of the shadow/highlights feature in CS2 and there are many outdoor photos with so much dynamic range that they are really helped by an appropriate amount of shadow/highlights.

But, I'm trying to understand exactly what the different settings in the dialog do. I know sort of what they do, but am interested in more technical detail.

Here's another good reference on shadow and highlight detail that I just came across: Part 1 and Part 2. This describes the amount of tonal information in shadows and highlights and discusses both using curves and the shadow/highlight dialog.

[rutt]

Fans of Dan Margulis should check out his explanation: http://www.dgrin.com/showpost.php?p=94240&postcount=1