PP5E Chapter 18 Summary - Part 1

This is the first discussion in a series of at least 3 on Chapter 18. Dan
covers a number of distinct but related topics here, some in more detail
than others. One is the Shadow/Highlight adjustment
(Image->Adjustments->Shadow/Highlight). I won't cover this here, but
instead refer to rutt's excellent tutorial on this:

Tutorial on Shadow/Highlight

rutt's tutorial takes you through Dan's approach to setting the S/H parameters
(3 of them for shadow,3 for highlight) which is similar to Dan's approach for
setting the Unsharp Mask parameters. The example in the tutorial is for highlight,
but it's the same for shadow.

In Chapter 18, Dan makes several points about S/H:

1. It is a great command, but underused
2. It's commonly believed a desparation play when really big moves
are necessary, but it's more useful where smaller moves are needed
(the picture in rutt's tutorial is in the latter category).
3. For big moves, other methods are better. He outlines another method
(the Bridge of Sighs in Venice), but doesn't go into a lot of detail.

I suspect Dan's thinking has evolved on this, based on a Q&A session in his March
class. I asked about Shadow/Highlight, and he used the time to go over the method
mentioned in #3.

In this discussion, I'll go through the blurred-inverted-overlay technique (which
Dan uses for the Bridge of Sighs) using as an example the picture rutt used in his
tutorial. I'll refer to this as the overlay technique for the purposes of this discussion.

The technique involves using a copy of the original on a layer in overlay mode,
applying an inverted channel to it, and blurring it. Let's break this down.

Overlay mode. This is what's done channel by channel:

- Anything 50% gray does nothing.
- Anything lighter than 50% gray lightens (the lighter, the lighter)
- Anything darker than 50% gray darkens (the darker, the darker)

If the layer in overlay mode has exactly the same image in all channels, then
the lightening/darkening of each channel is identical.

In rutt's example, we start with this picture:



rutt wants to bring out the detail in the highlights (waders primarily), so he
uses the Shadow/Highlight command to darken the highlights. With the overlay
technique, you want to find a channel where the things you want to darken are
very light (inverting this channel will make these things much darker, and so
using the inverted channel in overlay mode will darken these things).

Let's look at the three RBG channels.



The R channel is a winner. The waders are really light, along with the lower
part of the vest, and the face of the foreground fisherman.

So we start by making a copy of the background layer and changing its mode
to Overlay:



Then with the top layer selected, we use Image->Apply Image to put an inverted
copy of the R channel into all three channels of the top layer:



Note that we use Darken mode, since we just want to darken the highlights here.
With Darken mode, only the lighter parts of the R channel will have any effect
here (the darker parts of the inverted R channel).

The resulting picture looks a little odd, as do the intermediate pictures in
rutt's tutorial:



The next step is crucial. It's a Gaussian Blur that's similar in effect to the
Radius setting on Shadow/Highlight. To quote rutt's Shadow/Highlight tutorial:

We want to find a value that best brings out shape in the targeted areas.
Too low a value, and the highlights are blurred into a solid (as at the end of
the last step). Too high a value, and the the transitions are so sharp that fine
shape detail is lost.

In this picture, I found a Gaussian blur radius of 25.2 to be right:



The final step is to set the equivalent of the Amount in Shadow/Highlight - Opacity
of the Overlay layer. Start at 0% and set to taste. Here, I find a pretty high value
(78%) to be about right:



This is rutt's image with Shadow/Highlight:



In my opinion, the overlay technique does a better job here. Not only does it
bring out the waders at least as well as Shadow/Highlight, it also separates
the fisherman in the foreground nicely from the background. rutt limited the
Tonal Range in his tutorial to the waders, but I experimented with larger Tonal Ranges,
and the overlay technique did a better job in that case also.

This may seem complicated, but it really isn't. Most of it is mechanical. The
knobs are similar to those for Shadow/Highlight, but you have another important one -
choosing the channel that gives the best result, which Shadow/Highlight doesn't give
you.

To summarize the steps (to darken highlights):

- Create an additional layer that is a copy of the original
- Set the mode of the top layer to Overlay
- Find the channel where what you want to darken is light
- Apply Image to the top layer with that channel inverted darken mode
- Gaussian blur to get the detail right
- Opacity of the top layer to taste

A similar flow can be used to lighten shadows.

More work than the sliders on Shadow/Highlight, but not much.

This isn't a finished product. This is a starting point for other corrections. What
we've done is a first-order correction to get the tonality within reason.

The question is whether the minimal extra work is worth the effort, compared to
Shadow/Highlight. For this picture, I think it is. I'll be discussing this further
using pictures where I can't find a way to use the overlay technique that gives me
results even as good as Shadow/Highlights.

Some technical details, ideas:

- Sometimes you'll see a color change you don't like (it didn't happen in this
example). If that happens, just create a copy of the background layer
and put it right on top of the background layer, just below the layer
in Overlay mode. This layer has normal mode. Merge the top layer into
the middle one, and change the mode of the new top layer to
Luminosity.
- When I said that I put an inverted R channel into all 3 channels of the
layer in overlay mode, that wasn't exactly correct. Because I used darken
mode, it put the R channel into places only where the inverted R was darker
than what was there already (R lighter than what was there already).
This is OK in context. Any place where the the layer is different from the
inverted R channel, it is identical to the underlying layer, and has no
effect.
- You could use this separately for lighten or darken, or maybe find a channel
that does everything you want - what you want to lighten is dark, and what
you want to darken is light.
- There's no direct equivalent here of the Tonal Range slider on Shadow Highlight.
You could use a curve adjustment to the top layer to get the equivalent
effect. It's actually more powerful, as you can set the point where
lightening/darkening starts/stops, and control somewhat the amount of each.
This turns into a pretty complex workflow, and I'm not sure it's worth the
effort.
- Instead of using Darken mode, you could use a layer mask to limit the effect,
such as a K channel from a false conversion to CMYK (pathfinder's summary
of Chapter 5 inspired this idea). Similarly, you could use a Luminosity
mask (or an inverted one). I've just started to explore these.