(continued from Part 1)
1112 Good color, slightly washed out due to an overly light road. Comparing it to some others seems to me to indicate that there are advantages to having more detail in the sky than this one does.
1113 Speaking of which, now here is an interesting sky. It violates a basic principle of outdoor photography, which is that although clouds do get bluer as they get darker, it’s best to suppress that added blue for as long as reasonable. Here, however, it looks like the added blueness flatters the foreground. Overall the trees are slightly lighter than in some other version and I might compensate by retouching the flags to make their reds deeper.
Making this much use of the sky as a prop was also a feature of #1110 and especially #1101, not so much versions prepared by others. When people are thinking along the same general lines we often get the best result with a blend. So, here’s my suggestion. Grab #1101, #1110, and #1113. Try blending any one 50-50 into any other. I’ve tried, and I always find the blend better than either parent. (This would not happen if you blended one into a version that doesn’t play up the sky.) Furthermore, I’ve made an average of the three, called it par strong sky, and uploaded it to the folder. I’m not saying it’s better than the existing par but it’s a reasonable alternative.
1114 Chosen for the par version. Not much to criticize here. As it turned out, this one used nearly the same procedure as I did to create #1118, so I’ll reserve further comments until then.
1115 This is in many ways the most natural-looking submission of all. Fine definition everywhere, standout flags, good variation, sky interesting without being obtrusive. Paco Márquez got there with a simple, if drastic, procedure. He assigned a profile of Apple RGB, 1.0 gamma, and then multiplied. This gave him something quite conservative and too light, leaving him a lot of room to experiment in LAB.
The problem, if you want to call it that, is that the version is less colorful than almost all others. Having grown up in the area, I would say that the actual color of the leaves lies somewhere between Paco’s and the par version. The trees can be fairly red but not the red of a fire engine, or of a national flag for that matter. If this picture were simply for tourism to the region I would say Paco’s is perhaps the best version except maybe #1101. But since we are specifically told that the photo illustrates the most spectacular display of autumn color in half a century, the chances are we need more color than this.
Strategy: if you have a version that you suspect has a certain defect, such as possibly not being colorful enough, it is very easy to make another version that doesn’t have that defect. Starting from scratch, you should be able to quickly make something that looks a lot like the festive #1103. And then there are a million blending options. If the second version happens to be useful for anything other than blending with its color at a low opacity, so much the better.
1116 Another nice version, this one by Robert Wheeler. Good separation between the two most prominent trees, resulting from prudent use of H-K and MMM. He used Equalize to try to add contrast, a questionable move but it may have worked. I would, however, darken and intensify all the flags
1117 This person ran into a problem I referred to earlier. He submitted a different version, then a day or two later looked at it again and decided it was too loud, so he submitted what we see now. I would say these new colors are believable but the real problem with this version can be seen by comparing it to #1118. The main difference is that the lightest clouds in #1118 have been made much lighter than this, giving more range to the trees and resulting in a snappier rendition.
1118 Chosen for the par version. Here’s mine. Of the “standard” (non-heavy sky) versions, it has the most snap in the trees and the most variation in the critical areas. To see how that happened, I’ll compare it to the excellent #1114, because that person used very similar procedures. Some of the points are instructive so this discussion will be lengthy. I recommend opening both versions and toggling back and forth between them with Apply Image to note how they vary. Here are the things we both did:
*Philosophy. Here’s his explanation, with which I concur.
The original image looks like an afternoon autumn day with unsettled weather from sunshine
in the foreground to cloudy in the valley to rain or fog in the distant background.
So while there seems to be no doubt that the foreground should be sunny bright, the question
is how dark to leave background. I would think that the background wasn't as dark as
captured on film, and someone who was there to experience a once-in-a-generation autumn glory
wouldn't want a dark and gloomy area dragging down the memory. But neither should it appear
too light as if there were no rain clouds in the vicinity.
*We did limited color correction at the outset. Later, although we didn’t do things exactly in the same order, our workflows featured the following.
*We applied a false profile (he actually used an Exposure layer, which is the same thing) and masked the multiplying layer with the original blue channel, so as to retain detail in the sky.
*We both attacked the green channel, and then the red, trying to add contrast to them on Luminosity layers.
*We both used the Darken Sky action to get a heavier sky on the right side.
*We both felt that our shadows were plugging and opened them with a Hammer action and/or Shadows/Highlights.
*We both wanted to use H-K or some equivalent to cut down on excessive color.
*We both used forms of MMM to try for better variation in the reds and yellows.
*We both saw the need for heavy sharpening, although we did it in different ways.
Now, the slight tweaks that make the difference. In first looking at the picture we should all be thinking of trying to add tonal contrast to the trees. Some will necessarily get much lighter than at present. When this happens, we are in danger of severely damaging the sky and the flags.
We both realized that the sky is not an issue, it can easily be masked out because it is very light in the blue channel whereas all the trees are very dark. But the flags are not a problem either. They are just about detail-free in the original. We can restore their colors by hand later and no non-professional will be the wiser. So eventually I just painted the flags’ colors back in. Several other versions would have been improved with the same move.
Since I wasn’t worried about losing the light stuff, my first step was to plant a white point in the central clouds, you can see how this created a more interesting sky than in #1114. Then, on a new layer I applied a straight-line curve to the green channel, blowing out the highlights while maintaining the shadow point where it was. This lightened the green and the yellow trees significantly, the red ones not nearly as much. I then changed layer mode to Luminosity to avoid a grossly green look, and used Blend If to exclude all the non-tree stuff.
I now repeated this procedure with a second new layer, blending the RGB composite into the red, which had weak contrast originally, and again setting to Luminosity mode and excluding the non-trees.
In #1114 the guy did not lighten the clouds as much as I did. His moves to add contrast to the green and the red channels involved multiplying them into themselves. That works, too, but at this point his version would have been much darker than mine, which affected our next move.
We now each wanted to tone down all but the brightest reds, so that when we eventually added a color boost we would get some bright red accents in the trees but not overwhelm the viewer with redness. In principle, that called for H-K, which has one layer that adds grayness to all but the brightest colors and another that darkens them. He could not afford that second part because his version was already so dark. So, he used H-K Reversed, which lightens the strongly colored areas while desaturating the less colorful ones. This gave him the more subdued color he wanted in preparation for MMM, but it didn’t help contrast.
I could and did use that part of standard H-K, but I did not need the part that added grayness to the less saturated colors, because I had a way that’s subtly better. Go back to my first Luminosity layer, where I had lightened the green channel to add contrast. If that layer were in Normal mode, it would have been extremely green in the lighter trees, much less so in the red ones.
In moderation, that’s a good idea in this particular picture. If the idea is to emphasize reds without actually making them the same color as the flags, it’s better to force the non-reds toward green than towards gray. So I duplicated my Luminosity layer, changed it to Normal mode, and cut back opacity to 20%. Then, when I did a second Luminosity layer to darken the red channel (which added a cyan cast to the lighter trees), I did the same thing, making a Normal mode layer out of it at much lower opacity.
These two moves left me with a green-cyan cast in all trees other than the strongly red ones. Eventually, I found that the shadows were too cool, so I neutralized them with curves, but that cool cast is still present in the lighter half of the trees. So, the green trees are greener and the orange ones yellower. When trying to enhance reds, Chevreul suggests adding green elsewhere. Compare the green trees in #1118 to those of #1114 and see if you don’t agree with him.
1119 Like some others, this person tried a false profile approach, giving him flexibility to reassign contrast. He then assigned more of it to the sky and less to the trees themselves, resulting in a darker and flatter version than most, though the color in the lighter areas is fine. But the blues of the flags and the green of the Shuttle sign have gone black.
1120 This is most directly comparable to #1110, which is about the same darkness but retains color better in the shadows.
1121 Chosen for the par version. This attractive result was the result of two separate passes, each featuring MMM and one of the Hammer actions to add snap to the foliage.
1122 The par version.
1123 (posted, but not included in the download) the strong sky par version.