This exercise offers many opportunities for creative blending. It's shocking how often two or more mediocre versions can be combined to make one outstanding one.
Before Monday morning, I don't look at submissions carefully. The ones that are obviously lacking get set aside, and I make a tentative par version so that when deadline comes if a particularly good one happens to come in I can drop it in quickly. Once posted I evaluate and make notes on them in the order of numbering, which is random. So if I see that a certain version can be improved by blending with or comparing to another, that other image has to be one with a lower number (other than the par). So the higher the number you drew, the less likely you're going to be involved in a demonstration.
Some of these blending exercises produced the four numbered versions that I posted Tuesday. To view them you need to go to the Photos folder, or take a second download.
The group discussions of this exercise have been the best of any case study this year.
801 Off to an excellent start. This one should have been selected for the par version, as I now find it superior to some that were. It's also a great blending resource, as will be shown shortly. Reminder, though: don't delete any tangible objects. Here, some wiring and a couple of poles were deleted. They are small enough to probably not be noticeable, but for our purposes they make it harder to blend.
In not choosing it, I underestimated the value of the misty background in bringing out foreground contrast. I was influenced by a somewhat lackluster sky, and also by the unpleasant greenness of the background hills. That last is a valid criticism and I will show how to correct it in a demonstration after #813. Other mentions on this version are at #804, 808, and 809.
802 I suspect Gerald Bakker of having taken lessons from my wife. When something goes wrong, such as burning the salad, or submitting a version that is grossly lighter than anyone else's, it gets blamed on me.
The second step above is useful, and results in better overall color than many have. The first and third, however, are wrong, because they move the tonalities of the greenery and the pagodas closer together, whereas we would like to drive them further apart so that the structure shapes can be better appreciated.
In penance for having handed Gerald the rope with which to hang himself, I will turn this version into something that gives the par a run for its money, right after #803.
See additional comment about this version at #806.
803 We follow the lightest version with one of the darkest, by Christophe Potworowski, who is not pleased with it. As neither #802 nor #803 is going to be a popular choice, and as they represent two extremes, the obvious maneuver is to blend them. And indeed, a 50-50 blend with #802 produces something better than either.
That, however, is what we call a "stupid" blend, where no account is taken of the individual peculiarities of the parent versions. The par itself is a "stupid" blend, because it's simply an average of five parents, 20% weight each. Let's try something better, taking advantage of the fact that the most tonal contrast between the red and green objects in this image is to be found in the green channel.
Ingredients: #802 and #803.
1) Start with #802 and add a duplicate layer.
2) To it, apply #803, Normal mode, 100% opacity.
3) Oops! Cancel that! #803 is so vastly the darker of the two that it's easy to assume that it's darker everywhere. It's not. The sky in #802 has more detail, and is therefore the darker of the two.
4) Re-apply #803, this time in Darken mode. It's true that I could have simply changed the layer mode to Darken, but that would have left an inferior version on its own layer where it has nothing to do but make trouble. Also, I am not yet persuaded that we need to use any of #803's color, and would like to be able to change the mode of this layer to Luminosity to find out.
5) Merging the two layers is awkward because the pagodas of #803 are so dark. Lowering opacity to 50% won't work as well as loading a layer mask based on the combined green channel, where the pagodas are quite dark.
6) Blur the layer mask with 30 pixel Radius or so. This step is always necessary when blending two layers of very different darknesses.
7) Change layer mode to Luminosity and decide whether it is now better. My opinion is that it is not and the step should be cancelled.
8) We still would like better tonal separation between grass and pagoda. So, add a curves adjustment layer. Open the green curve (where the difference between the two is the greatest); place one point to hold the grass, and darken a point that corresponds to the pagodas.
9) Change mode of the curves layer to Luminosity.
The result is #836, which two people have remarked is better than the par.
As my late mother used to say when I was rude to her, "Isn't it a miracle that two such stupid persons as your father and myself could have produced such a brilliant child?" Maybe if she had seen this demonstration she wouldn't have been so sarcastic.
804 One of the most subdued versions. The person says:
Well, there's nothing wrong with dull coloring, some people like it that way, although it seems to me that a more colorful sky would be helpful. The big problem is the lack of break between the pagodas and their surroundings. The easiest way to prove this is to blend #801 into this in Luminosity mode, which makes the pagodas jump to the foreground. A more complete demonstration, however, is shown at #807. Demonstration of how its subdued colors are useful in other blends are at #811 and #815.
805 A pretty, and somber version, from a student of Matthew Croxton.
806 The individual explains this tastefully understated version:
He did build up the sky, but not to the circus-like colors of some others. And the color answers the criticism of those who feel the par is too gaudy. To prove it, put this one as a layer on top of the par in Color mode. I don't know that we'd keep the opacity at 100%. But 50% is an improvement IMHO, and I think 25% would get a unanimous vote of approval. So, arguably this version has better color than the par.
The luminosity, OTOH, is another matter, because this person blended with the red channel, Darken mode. His reasoning was to darken the background hills, which I'm not sure is a good idea. What is a bad idea is that it also darkened the greenery. This was essentially the same mistake made in #602.
Ingredients: #806 and the par.
1) Start with #806 and add a layer.
2) To it, apply the par.
3) As we have seen, the color of the par may be no better, but the contrast surely is, so change layer mode to Luminosity.
4) Reduce opacity to taste. I chose 50%.
5) Time now to find out whether any of the par's color is useful. So, duplicate the par layer and set it to Color mode, 100% opacity.
6) Evaluate. I say that the par color is better in the sky, but not so much elsewhere and definitely overbearing in the pagodas. So, add a layer mask to the par Color layer, and load the RGB composite into it. No need to blur the mask, since there is no luminosity change as a result of the merge.
The result is alternate par version #839. #806, because of its agreeable color, is also used for further demonstrations in #825 and #828.
807 Chosen for the par version, rejected, and then rechosen. This was likely the one that Ken Harris was talking about:
Ken and I don't share notes so I don't know for sure this is the one, but I can say it was in my initial group until a bunch of good entries came in on the last day. So I replaced it in the par stack, only to discover that although the replacement was a better version, it was also a darker one. The par became worse, lacking snap. So I reversed course and put it back in.
Here's Bill Theis's take on his version:
To summarize: the color here isn't wrong, but it's overstated. The luminosity part, however, is quite good. It's basically the opposite, then, of #806, which had nice subdued colors but not enough contrast.
However that's not the one I'd pick for the blend. Since the object is more subdued colors, I'd go for a blander one provided that its coloring was basically correct. We have such a candidate in #804.
The group discussions here have been the best of any case study this year. :
Ingredients: #804 and #807.
1) Start with #804 and add a new layer.
2) To it, apply #807.
3) Change mode to Luminosity and decide whether it is now better and if so, what opacity to use. I chose 70%. The reason I do not keep this layer in Normal mode is that I am sure we want some of #807's color, but not as much as of its luminosity.
4) Duplicate the layer, change its mode to Color, and again decide on the opacity, if any. I think it that #804's color is so dull that we should use some of #807, but I chose only 35% opacity.
5) We still would like better tonal separation between grass and pagoda. So, add a curves adjustment layer. Open the green curve (where the difference between the two is the greatest); place one point to hold the grass, and darken a point that corresponds to the pagodas.
6) Change mode of the curves layer to Luminosity.
The result is alternate par version #838.
808 This person made a mistake similar to others in that he blended into the green channel using the red channel as a source. This unfortunately darkened the grass with respect to the pagodas. The result is similar to (although more colorful than) #804 in that #801 in Luminosity mode will improve it.
809 By my count, this makes four out of the first nine versions that would benefit substantially by blending with #801/Luminosity.
810 The #801 blend might not help here because the presentation is so moody. John Castronovo:
811 Edward Bateman:
The sky and the pagodas are fine. I have no objection to them being so dark. Granted that they are dull, the saturation of the grass seems out of place. The blue-green background hills are very distracting. So I'd wheel out the dependable #804 again to desaturate them as follows.
Ingredients: #804 and #811.
1) Start with #811. Add a duplicate layer.
2) To it, apply #811. The character is so different that there is no point in thinking about a luminosity blend. Just change the mode to Color.
3) Move the extremely dull result into LAB, without flattening.
4) Go to Blend If, and exclude anything that's A-positive on the Underlying Layer. That brings back the pagodas, fhe flowers, and the sky. The only things being dulled down are the greenery and the hills.
5) Reduce opacity to taste. I chose 65%.
The greens are only slightly duller now, but one can really see how much the blue-green hills were distracting from the pagodas.
812 I referred to this one in a message to Doug Schafer in another thread. I originally had this as a par contender, largely because it has IMHO the best-looking sky of any entrant. The foreground, OTOH, is good but not great. The pagodas themselves are fine, but everything else is too blue, especially the background hills.
Fortunately in this image it's easy to make use of this sky in other contexts. I show examples in #825 and #828.
813 Not one of the better versions overall because of the lack of tonal contrast between grass and pagodas, but it has its uses. The foreground is somewhat biased toward yellow, which pushed the background hills away from blue and toward green. The person did not like that, so he nudged them toward magenta with Hue/Saturation.
This was a good move and should probably have been done in #801. You remember that one: excellent foreground contrast, slightly downgraded for the greenness of the background. It can now get the minor tweak that it needs.
Ingredients: #801 and #813
1) Start with #801 and add a new layer.
2) Apply #813 to it. The contrast is obviously much worse, so we'd ordinarily immediately shift to Color mode. Here, however, we have to figure out how to merge the two layers, and it's easier to visualize that by seeing the detail.
3) Add a layer mask. To it, apply the blue channel from the base layer. That's chosen because of all available channels, it has the biggest gap between the hills and the pagodas. No need to blur this mask since there isn't going to be any luminosity change.
4) Now, finally, change mode to Color, and decide whether to reduce opacity. I saw no need to and left it at 100%.
The background is now warmer, the hills are on the magenta rather than the green side of blue, and #801 is presented more effectively.
814 Could not be considered for par version or blending due to changed aspect/lens profile. Otherwise it would be nice to compare it to #804.
815 Chosen for the par version. David Remington:
I like it too, it has the richest color of any of the versions with dark pagodas. I'd say that granted the darkness they're a little too fiery, but that's not a big deal. What is a big deal is that the background hills are dark enough to distract from the pagodas. Worse, the heavy cloud in the upper right is completely counterproductive. It takes away from our appreciation of the pagoda towering against the sky. This is not a matter of taste, IMHO, it is an outright error. To correct, we need something bland, and we know what that one is.
Ingredients: #804 (again) and #815.
1) Start with #815 and add a duplicate layer.
2) To it, apply #804 in Lighter Color mode. Since the foreground of #816 is already lighter, not much changes except the dark cloud and the mountains.
3) Change layer mode to Luminosity so that the sky color doesn't fade.
4) The heavy lifting is already done, but just for the sport of it duplicate the #804 layer and change mode to Color.
5) This desaturates everything, whereas we'd like to confine the effect to the pagodas and possibly the grass. So, add a layer mask and apply the red channel from the background layer, inverted. That will permit some desaturation of the pagodas, a slight amount to the greenery, and very little to the sky.
The result is alternate par #835. I'd suggest toggling the middle layer off and on to see the huge difference it makes.
816 John Furnes:
It came out reasonably well. To me, the desaturation of the sky and background hills is somewhat jarring, but the pagoda shapes are still well defined.
817 This interesting treatment could not be considered for par or for blending due to a change in aspect. John Gillespie:
Like some others I went for a darker, somber look based on how the lighting appears. But it is a very difficult image to make attractive. To paraphrase, what is interesting is not believable and what is believable is not interesting. Taken from a different angle the pagodas could be rendered as a silhouette against a dramatic sky, but they take up too much of the foreground for this to work.
I don't know about that part. I agree that they shouldn't be fully black, but I have no objection to making them so dark that a certain amount of definition is lost. Several versions, including my own, do that. It's effective because the pagodas are recognizable by their shape.
818 Chosen for the par version, mostly for its beautiful texture and interesting orange sky. Robin Mark D'Rozario notes that his version is warmer than average. It also has lighter pagodas and darker greenery than the par does, making it less contrasty than it should be. Since #817 doesn't have that particular problem, IMHO blending 20% of it into #818 improves things.
This version is featured in another blend at #829.
819 Too heavy-handed in the shadows. Greenery is black when the pagodas are lit.
820 Full of foreboding, this person wrote in his procedure summary,
When an image has a cast, as this one does, we normally don't see it unless we have something else to compare it to, like 38 other versions that aren't as green as this one. But look at it by itself for a while and the cast vanishes because the eye adjusts to it. See Edward Bateman's comment at #811.
The solution is to check known colors. The sky was obviously treated separately and is irrelevant. The vegetation is garish but not wrong by the numbers. But the pagodas? They should be a dull red, nominally A=B. To find A=0 and B positive is clearly wrong no matter where in the pagoda it's taken. Worse, the background hills are typically (10)a(5)b--more green than they are blue. That can't possibly be right.
See #821 for an instant solution to these issues, and #822 for another comment on this version.
821 A purple cast, detected by measuring the pagodas, which are sharply magenta. Also, this person made the same mistake as in #806 of using the red channel as the source for a luminosity blend. As there, this brought the distant mountains closer to us, and also darkened the vegetation to the point that it interferes with our perception of the pagodas.
Back to the cast, though. I just said in #820 that the sort of brick red found in these pagodas can normally be expected to be approximately A=B. In this version, a point in the light area at bottom right of the main pagoda, just above the flowers, measures 43L19a10b. The same point in #820 measures 39L20a35b. One therefore has a magenta cast and the other a green one. Since magenta and green are complementaries, just blend the two together at 50-50 or to taste, and the result is much better than either. Although this is not a politically correct way to proceed, these two casts have just encountered cancel culture.
822 Similar to #820 in that the background hills measure a disagreeable green, with the added complication that the sky is less pronounced than most people like.
823 Kent Sutorius before:
Attached is the Land of Pagodas submission. For me this was by far the most difficult image to work on. Almost didn't submit anything. I tried different things in the RAW, flat and default images with poor results. I think I am in the danger zone now that I know how to do a number of things but not the experience/knowledge to know when to apply in the workflow. I would get portions of the image great but other portions wrong. Had problems with the mountain range. It kept getting washed out when the rest of the picture in the foreground looked good. I ended up doing simple adjustments: Levels, Select Color, Curves, and high pass sharpening.
Also the overly blue hills are a defect. I hope the other examples make clear how the foreground contrast could be corrected.
824 Jorge Parra:
Well, I wouldn't reduce this haze because it makes it easier to perceive the smaller pagodas in front of it. With respect to this picture, I'd go through a procedure similar to what was demonstrated at #815.
825 Steve Jenkins:
I don't agree that nothing about this looks real; the foreground is too yellow but no more unreasonable than a dozen others. The real problem is that the sky is ridiculous, and there is an ugly transition line where it meets the hills. Apply 50% of the color of #806, pop in the excellent sky from #812, and all is well.
826 Edward Bateman's students did well with the Choir image but apparently tried to hard to knock the blueness out of the background hills and made them yellow instead.
827 From previous comments, I imagine you have surmised that I prefer a very dark version. When I prepared this one, I fully expected it to be the darkest of any reasonable entrant. I have no objection to it being a semi-silhouette of the pagodas. I for some reason decided that such a dark foreground demanded a brilliant sky and then produced one that was technically just about as clever as the idea itself. If I drop in the sky from #812 there's a huge improvement.
Like some other people I painted color into the smaller pagodas in the background, trying to make them more prominent.
Whether you like this one or not it's a useful auxiliary file. If you feel that your own file is somewhat light, blending this thing in at 25% should
828 Here is another lesson about why not to be too quick to reject something that doesn't appeal on first glance. The photograph has at least three items that can be treated separately: the sky, the overall color, and the overall contrast. This version fails in the first two categories, but is reasonably good in the third. So, let’s put together some familiar blend sources.
Ingredients: #806, #812, #827, #828
1) Start with #828.
2) Add three new layers.
3) To the first new layer (the one right above the background), apply #806.
4) Change the layer mode to Color, and make a judgment as to what improvement, if any, it makes. Having decided that, change the mode to Luminosity and make the same judgment. When this process is over, you'll know how to proceed.
I say it's pretty easy in this case. I see #806 as absolutely and unconditionally better than #828 with respect to color, and absolutely worse with respect to luminosity. So, I set the layer to Color mode, 100% opacity.
5) To the second new layer, apply #812, with its excellent sky.
6) Unless you've decided you want to make some use of its foreground, and I don't, go to Blend If, Gray, Underlying Layer. As the sky is lighter than any other part of the image, the shadow slider can be moved to the right until the entire foreground is restored, but the new sky is retained. When satisfied with the positioning of the slider, Option-click it to break it in two, and move the two halves slightly apart. This prevents a harsh transition where the two layers meet.
7) Decide whether to retain this layer at 100% opacity. I liked the effect at 67%.
8) Since what we now have is good, but it doesn't happen to be one of the darker entries, follow what I recommended in #827: apply it to the top layer, and set opacity to 25%.
IMHO this is now a serious competitor to any of the par versions, although it wasn't finished in time to be posted with the other alternate pars.
829 Bill Iverson:
The problem is that they aren't reddish brown, they're magenta, with the A value considerably higher than the B. The other issue, which is more debatable, is that the background is quite a bit lighter than what most of us prefer. The contrast in the foreground is nice, however. Limiting myself to versions already seen, I would pick up #818. Applying it at 60% Darken gives a better sky and also to some extent corrects the color.
830 Hector Davila did not fall into the trap of blending the red into the green and thus bringing the grass closer to the pagodas. Instead, he did the blend in reverse, resulting in the lightest greenery of anyone and a huge contrast to the pagodas. The color is kind of lemon-yellow, not so good by itself but certainly usable in a blend. Because this is an excellent auxiliary file, a companion to #827. One or the other almost has to improve most versions. Therefore, one final
Ingredients: #827, #830, and your file.
1) Start with your file and add two layers.
2) Apply #827 to one and #830 to the other.
3) Change opacity of each layer to 25%
4) Toggle the layers off so you see your own version again.
5) Toggle one layer back on and make a judgment.
6) Then the same with the other layer.
Now, whatever you may think of #827 and #830 individually, you'll have to admit that they showcase the pagodas. So my prediction is that one of these layers is going to improve your file and the other will hurt it, but which one will help it depends on the character of your version.
831 Kent Southers:
I don't understand why there should be an issue where the structure meets the sky, Blend If can solve that problem even in RGB. The bigger issue in the sky is a repeat of the mistake made in #815 of having the cloud in upper right so dark that it diverts attention from where it belongs. Furthermore, the background mountains are dark enough to disguise the shape of the pagodas. The following simple procedure should demonstrate how important these things are. As with #825, we need to locate a version with a lighter sky but darker pagodas. In that one I used #824, but a better candidate has since reared its ugly head.
Ingredients: #827 and #831.
1) Starting with #831, add a duplicate layer.
2) Apply #827 to it, Lighten mode. As the foreground of #827 is darker, it's unchanged. The sky and the background get lighter, however.
3) Reduce layer opacity to taste. I say 70%. Now toggle the layer on and off to see the impact. According to me it's effective, but we aren't quite done.
4) Change layer mode to Luminosity. I don't think there's much of a way to know in advance whether Luminosity or Normal gets a better result, so trial and error is called for. Here it’s a close call. I have a slight preference for leaving the layer in Normal mode.
832 Chosen for the par version, as predicted by James Gray. Ken Harris's summary:
They didn't seem to find any way to improve on it, except possibly the next entrant.
833 Chosen for the par version. The best has been saved for last. This is my personal favorite. Excellent, deep brown pagoda color, sky interesting without being overwhelming. Great illusion of an endless valley of pagodas. In all of these respects it's approximately equivalent to #832. The attractive dull color of the pagodas may have been the result of H-K; however to me the best touch is some dodging and burning to the sides of the foreground pagodas to create the illusion of reflections.
834 The much-maligned par version.
835 The blend described at #815.
836 The blend described at #803.
837 Prepared by John Castronovo by altering the par version only.
838 The blend described at #807.
839 The blend described at #806