Shasta: Dan's comments


Dan Margulis
 

Anyone who’s done forensic work, such as enhancing pictures of crime scenes to bring out evidence to be used in court, might find this one somewhat familiar. The object of interest is almost nonexistent in the original capture; our job is to make it visible and understandable.

The difference is, if this were an enhancement of a bloody footprint, say, the police wouldn’t care whether the footprint is at all accurate for color and would care even less whether the rest of the image was attractive. Our job is harder because we do have to care about both these things.

You might consider that this is similar to the choir itself in the Choir exercise, or the beach itself in the Beach at Sunset. In each case we had to make major changes, though admittedly not of the magnitude of the ones the mountain needs.

The similarity breaks down there. If you decide that you want to borrow somebody else’s choir or beach and paste them directly into your version, well, first of all the selection/masking would be difficult. But even if you could merge them gracefully, it would still be obvious that something was wrong, because these items are closely related to the rest of the scene and we could see that the pasted-in elements don’t quite belong there.

In the current exercise, that’s not true. The background is almost 100 km distant from the foreground. They share no lighting. They do not have to match up in any way. We can take the background from one picture and slap it unchanged into another, and no one will be the wiser. Furthermore, doing so is a snap: you can drop in a new background in a couple of seconds with Blend If in LAB.

So, you have two halves of the picture, interchangeable with other versions. The background needs drastic, horrendous changes. The foreground is basically OK as is. We have already noted the artistic choice between treating the mountain as a realistic object or as a mysterious-looking one. A more fundamental choice, however, is this: do you

1) work on the image as a whole, trying mightily to prevent damage to the foreground as you do what you must to the background, or

2) do separate versions for the foreground and the background and only merge them at the last minute?

To me, this seems like a very easy choice. I did it the second way and so did at least one other person, I assume that the reason most did not is that they didn’t think of it because it’s rare to find an image where it’s even possible.

If there were no mountain in the background we wouldn’t need to devote a case study to photo. The foreground is simple in PPW or for that matter any competent method. The background isn’t so easy, but if you’re allowed to let the foreground become nearly black it’s not so hard either. The biggest issue would be artifacting.

Ken Harris remarked a few minutes after the versions were posted that he could easily pick out the four best versions. Looking at the scene overall, I think I agree. But I wanted to go further. Suppose we treat this as two separate scenes. Who has the best foreground? Who the best background?

If I wanted to find out I'd quickly go through each version and discard the ones that clearly aren’t competitive. That’s what I did for the 28 versions other than the two pars.

Looking at foreground only, I discarded eight versions immediately. The other 20 would require a second look before narrowing the field down further.

Looking at background only, I found eight reasonably well-structured mountains. Five others weren’t really so pronounced, but the treatment was interesting. Those five became the “mysterious par”. As for the other 15 backgrounds, the less said, the better.

All you have to do to make the foreground acceptable is to increase saturation slightly, because its hues are basically correct at the start. Given that, it isn’t encouraging to have as many as eight unacceptable foregrounds. Some of these, I know, happened because people were trying too hard to fix the background and the foreground paid the price. Of the eight versions I rated as having acceptable backgrounds, three had unacceptable foregrounds. No coincidence.

Those who did get a nice background used various tools, many of which are found outside Photoshop. Boosting darkness of blues in acquiring the image; Simple curves, multiplications, the Velvet Hammer, the MMM actions, luminosity blends featuring the red channel, and one person even selected the mountain and hit Equalize. Often people would use a combination of several, which is better than repeating the same move every time because every method produces a different artifact patter. Notice that there are no obvious artifacts in the par version. There were in some of its parents, but averaging different kinds of artifacts lessens their impact.

All of the above may work, but there are Wrong Ways. First, if ever there was an image that requires a blown-out highlight, this is it. The snow is white, so it should be around 0a0b. The reason we normally set highlights at darker than 100L is that we lose detail otherwise. But in the lightest snow, which can be located with a Threshold layer on the mountain’s left side, there isn’t any detail to protect. So that particular area should be, in principle, 100L0a0b. Surely it should be no darker than 98L. Yet 16 of the 28 non-par versions did, their highlight was 97L or darker. I’ll show the impact with demonstrations tomorrow, and tag every image with its highlight value.

It is also wrong to use the Bigger Hammer action (the less violent Lesser or Velvet Hammers are probably OK). The Bigger Hammer is a desperation measure when you must have additional highlight detail at all costs. That is the case in a shot of Niagara Falls, when we’re desperate for more realistic spray. It is not the case here, because the snow has no inherent detail to speak of. The Bigger Hammer will emphasize detail quite efficiently—in the artifacts.

Some also used multiplication layers, which are fine for bringing out depth in the mountain. From force of habit, though, some people started with a false profile. No need for thatif you are preparing one version for the sky and one for the background.

Finally, it’s nice to have a certain amount of blueness in the sky to set off the whiteness in the snow. The top right of the sky is large and dark, and more than sufficient to do this. More blue than that amounts to beating a dead horse. The whiter the clouds are, the better. Also, we know that the front of the mountain is shrouded in smoke, but we would just as soon that the viewer not know it. The smoke is likely slightly blue in real life. We don’t want to emphasize that.

More tomorrow, with some demonstrations.

Dan





James Gray
 

My version is an outlier.  By me saying that I am sure most of you who look will know which image is mine without me giving its number.  I will also say it is ok with me if Dan or any of you are not gentle in the critique of my version.  I think what I have included from Dan's comments are relevant to what I thought about the image and what I tried to do.  As Dan suggests the background needs drastic, horrendous changes.  I made drastic changes to the background.  In my first comment in the steps I sent to Dan when I submitted my version I said:

I assume that the final image should resemble what you expected to see and would have seen without all of the smoke and not just be a slight improvement of what was seen.

In the middle of my steps I wrote:

Clearly need masks to work on the sky and mountain separately from the foreground

So, I knew as I was working that the background and foreground needed to be worked on as separate images and then combined. I like my version of the mountain better than any of the other versions including the par. I am sure that most will have reservations about my dark blue sky, but I kind of like it. I think my most important step in improving the mountain was the haze slider in ACR. Clearly I let my drastic changes that improved the mountain affect the foreground. In other words, I did not completely follow my own notion of working on the foreground and background separately. However, in several of my steps, I did work on the foreground and background separately. One mistake I made is to not recognize the foreground was basically OK. I did too much to it and then tried to backtrack using various blends. I should have started over creating a more modest foreground. I have not attempted to blend my version with any of the others, at least no yet. In case anyone has doubts mine is #925.

Jim Gray

On Wed, Mar 31, 2021 at 11:58 AM Dan Margulis via groups.io <dmargulis=aol.com@groups.io> wrote:

 if this were an enhancement of a bloody footprint, say, the police wouldn’t care whether the footprint is at all accurate for color and would care even less whether the rest of the image was attractive. Our job is harder because we do have to care about both these things.

 The background is almost 100 km distant from the foreground. They share no lighting. They do not have to match up in any way. We can take the background from one picture and slap it unchanged into another, and no one will be the wiser. Furthermore, doing so is a snap: you can drop in a new background in a couple of seconds with Blend If in LAB.

So, you have two halves of the picture, interchangeable with other versions. The background needs drastic, horrendous changes. The foreground is basically OK as is. We have already noted the artistic choice between treating the mountain as a realistic object or as a mysterious-looking one. A more fundamental choice, however, is this: do you

1) work on the image as a whole, trying mightily to prevent damage to the foreground as you do what you must to the background, or

2) do separate versions for the foreground and the background and only merge them at the last minute?

To me, this seems like a very easy choice. I did it the second way and so did at least one other person, I assume that the reason most did not is that they didn’t think of it because it’s rare to find an image where it’s even possible.

If there were no mountain in the background we wouldn’t need to devote a case study to photo. The foreground is simple in PPW or for that matter any competent method. The background isn’t so easy, but if you’re allowed to let the foreground become nearly black it’s not so hard either. The biggest issue would be artifacting.



Dan Margulis
 

Highlight values are given for each version, measured in the lightest area of the snow, sampler set at 5x5 pixels. Since there is almost no detail in the snow, the desired value is 100L0a0b. The measured values are highlighted when L<98 or when the A and/or B values are negative.

I have produced five alternate par files from the demonstrations below, plus an auxiliary file that can be used for adding definition to the background. These were not included in the original package, so you would need to grab them from the folder in the Photos section, or download all six in our Files section at
040121_Shasta_alt-pars.zip

Dan
***********

901 (100L 0a 1b) Chosen for the par version. Look who the cat dragged in to kick off the study!

I was done with this in about fifteen minutes. I saw no reason to work on the entire file at once given the enormous difference between the two halves. So I did one version for the background, a second for the foreground, and merged them together with a Blend If.

Most of that time was spent on the background, as might be expected. Noting the artifacts, I combined several approaches to bring out the mountain: luminosity blending, curves, Velvet Hammer, multiplication, plus some moves in LAB. Why so many moves? Because each creates slightly different artifacts and they tend to cancel one another out.

Like several others, I found that the bottom half of the background was too blue, so I made a desaturation layer and toned it down; a description of the method is in the demonstration at #914. I wish now that I had gone farther with it and also lightened the front of the mountain more.

Having finished the background, I turned to the foreground and did a PPW version in short order, followed by merging the two.

The problem, in retrospect, was that instead of two versions I should have had three. It seems to me that this mountain is too much of a good thing. Because it is so strong, I felt I had to make a foreground that's darker than most of you did. 

I should have learned from how we work with Color Boost. The theory behind that is that it's easier to have too much color and back off than it is to try to force color in until there's enough. So I should have been more conservative with my first try at the background, and then had a stronger background file in reserve in case I later decided that the mountain needs to be more pronounced. Too late for that now, but I now have posted such an over-the-top file as #931, and will demonstrate how to do it and what use to make of it at #909

902 (94L 0a 1b) The person's submission notes say:

This was a good exercise. I was actually a little surprised by how much detail of the mountain I was able to bring out. In the past, I'm sure I've just rejected at least a few images like this one. I'll be more likely to give them a go from now on.

The result is indeed impressive when compared to the original, but not when compared to some of our better versions. For now, I'd like to show only the importance of setting a white point, since the numbers suggest that parts of the snow are too dark here.

Demonstration:
Ingredients: #901 and #902.
1) Start with #902. Apply to it, #901 in Lighten mode.
2) Toggle back and forth to see the big improvement.

Explanation: #901 has a far darker mountain overall than #902 does, except that, as the number show, it has a real white point and #902 doesn't. So, that part of the snow in fact is lighter in #901 and the blend permits it. Also the center of the foreground gets lighter, which is also desirable. 

903 (99L 2a 14b) Chosen for the mysterious par; the yellowness engineered into the reflections on the summit is intriguing, as is the cyan flavor in the sky. This person decided he wanted to match the sky of the reference image. Instead of producing two versions and merging them, however, he relied on fancy masking to protect the foreground. It didn't quite work. No matter. Assuming that you like this background, it's easy enough to put more pep in the foreground if a suitable other version can be found.

Demonstration:
Ingredients: #902 and #903.
1) Start with #903 and add a new layer.
2) Apply #902 to it in Normal mode.
3) Since #902's background is so light, it's easy to use Blend If to exclude it. Choose the blue channel, This Layer, and knock out anything that's light.
4) Now that the foreground is the only thing being changed, reduce opacity to taste, and decide whether to use Normal or Luminosity as the layer mode. I decided on 50% Normal mode. This definitely peps up the foreground without altering the background.

904 (99L 0a 0b) Chosen for the par version. Similar to #901 with better trees, it's my personal favorite. This individual, who was not using Photoshop, went with two versions all the way, using a mask based on the B channel to blend them rather than Blend If. See another use for this image in #905.

905 (96L 0a (4)b) Good detail in the mountain, but too much blueness in the background but a poor foreground. This person used Equalize for the mountain, that accounts for the blue and for the extra blue, but it bought him a lot of retouching to get rid of the resulting artifacts.

To see how it would look with a real white point and somewhat less blueness around the mountain, blend in #904 in Lighten mode.

906 (98L 0a (1)b) These numbers suggest that things might be a little green. Do you see it?

Robert Wheeler: 
I blended the red channel to get more mountain detail, but found that ended up less than ideal (or I was not addressing the consequences adequately). Made a CYMK copy and found the cyan channel had even better mountain detail, so used that early in the process. A camera raw filter using de-haze helped the background more than I expected. Moving to Lab color space and steepening the b curve (a bit more on the blue end than the yellow end) helped intensify yellows and blue/cyans; had to leave the a curve untouched as my experiments with it mostly made too much magenta. I suppose having the mountain a bit darker like PAR would have improved it, but I do like having the dried grass paler yellow with less red/magenta tone because that is how I am used to seeing such patches in reality. Kate Wolf had a refrain in one of her songs about the rolling golden hills of California, referring to the color of the dried grass.

A comment about how to improve this version is at #907, and a demonstration at #911.

907 (97L 2a (1)b) This person used the Darken Sky action to get more depth in the mountain, but not quite enough. It is interesting to toggle between #906 and #907. The measurements show that one has a highlight that's slightly green and the other slightly purple. That certainly seems to carry over into the picture as a whole.

Since the two casts are opposites, they would cancel one another. #906 is probably preferable for detail. The prediction would be that blending 50% of #907 into it in Color mode would improve things. It does.

908 (100L 0a 0b) Chosen for the mysterious par. Nice white highlight, snowy appearance. No sharpening.

909 (94L (2)a 2b) Chosen for the mysterious par. A pleasant misty look but it would be better with a white highlight. Given the choice to have a relatively gray sky and the greenness of the measured highlight, it might be worth trying a warming filter in the light areas, to get somewhat of the effect found in #903.

This person's High Andes entry is already in, and he commented as follows in his process notes:

I am learning so much by these comparisons and I realize that I still have big problems. For example, in the previous image (Shasta) I have not been able to understand how others have been able to add so much to the structure of the mountain without creating artifacts in the sky.

Like this:

Demonstration:
Ingredients: #909 and a copy of the default version.
1) Move the default version into LAB.
2) Blur the A and B channels at two different Radii, I chose 1 and 3 pixels, this to prevent the appearance of brilliant noise in what follows.
3) Make a duplicate layer and set mode to Multiply.
4) Duplicate this layer twice, making three Multiply layers in all.
5) Add a curves adjustment layer.
6) In the Lightness curve, bring the white point toward the center to lighten the highlight. I chose 90>100. And darken the midtone; I chose 16>46.
7) To the attached layer mask, apply the newly darkened Lightness channel.
8) Blur the layer mask at, say, 60 pixel Radius. The mountain starts to take shape.
9) Duplicate the adjustment layer and its mask, doubling the move.
10) Make a composite layer. I
11) To it, apply the Surface Blur filter to reduce noise. I chose Radius 5 pixels, Threshold 9 levels.
12) Reduce opacity of the Surface Blur layer to 50% to reduce the "plastic" look.
13) Flatten and convert to RGB.  

This has produced #931, the auxiliary file that will strengthen the sky.

14) Now back to #909 and add a new layer.
15) To it, apply the newly created #931.
16) Add a layer mask, and to it apply the blue channel. In #931 the foreground is almost completely dark, so nothing will be allowed to change. Most of the impact will come in the mountain. The darker blues of the sky will be somewhat excluded, so that, due to the mask, they'll be lighter in the final result.
17) Blur the mask at 90 pixel Radius. This strongly brings out the cloud pattern.
18) At this point many options exist: lower opacity, changing the blue to a slightly different hue, etc. I did none of this. The result is posted as #932.

910 (89L 5a (3)b) Good foreground, background hampered by a very dark highlight plus banding in the sky, reason unclear from the explanation.

911 (91L (1)a 2b)  The individual writes,

I got this 95% to where I wanted to be in capture 1 in five minutes (layer RT), then, because I should be doing three sets of taxes, I spent an hour wandering around in photoshop revisiting old methods despite knowing the likely answer, given that I had an opinion about how the light should feel.  Although it’s not an especially good shot of things, I thought it could be made into a shot with a special feeling about the disappearing light.  I’m excited to see how other people treat this one.

This result here is neither fish nor fowl. It doesn't make the mountain pronounced enough to compete with the par but it isn't as interesting as those chosen for the mysterious par. One of the main reasons IMHO is that the lower clouds surrounding the mountain are too blue. Neutralizing them gives a chillier, moodier effect, especially if the white point can be set at the same time. For example,

Demonstration:
Ingredients: #906 and #911.
1) Start with #911 and add a duplicate layer.
2) To it, apply #906 in Lighten mode. That mode is needed because otherwise the trees would get too dark.
3) It unfortunately also makes the foreground golden field much lighter. Here's one of the few times where Blend If works reasonably well in RGB: just use it to exclude everything fairly dark on the Underlying Layer.

I think this, saved as #933, sets the mood a lot better. Other comments about this version are at #914 and #921.

912 (100L 0a 0b) This person got quite frustrated trying to color-correct the image as a whole and wound up with an overly blue foreground.

Demonstration:
Ingredients: #909 and #912
1) Start with #912 and add a duplicate layer.
2) Image: Adjustments>Desaturate, turning it into a grayscale.
3) Load as a layer mask, the RGB composite. This will restore the blue at the top of the image but desaturate much of the rest.
4) Since we don't want the mountain to seem grayscale (just less blue overall), reduce opacity of the Desaturation layer to taste. I chose 65%.
5) Add a new layer and apply #909 to it.
6) Do the reverse of the Blend If described in the #911 demonstration.

The result is posted as #934.

913 (100L 0a 0b) Too yellow in foreground, artifacting in the upper sky.

914 (91L 1a 4b) David Remington:
The tricky part for me with this image was building up Mt Shasta without overly affecting the sky which is prone to posterization. I wanted my image to be "believable". I thought my foreground was okay, but comparing it to the par, and its lineage, I see that it is a bit flat, maybe a bit green. Layered with the par in luminosity mode anywhere from 40 to 60% adds snap and definition. 30 to 50% color mode takes some green out of my foreground and cyan out of my sky. I find these changes are an improvement.

Thinking about the sky/mountain separation, I think color helps. I find the par's Shasta a bit stark though. Looking back at my Camera Raw adjustments only exported file, Shasta and the sky start out at the same hue with very little color in the mountain. Where to go from there.

Auto Tone might be a good start. This highlight is too dark.

I agree that the foreground color is a bit on the green side but we certainly have worse. There's good separation between the yellow flowers in front and the rest. IMHO the biggest problem is not that the background is the wrong kind of blue but that there is so much of it as to become overwhelming. The smoke in front of the mountain is probably somewhat blue in real life but we are not anxious to emphasize this. This is the same theme as in my demonstrations at #911 and #912, but here the mountain is more pronounced so the advice goes double. And there's no need to steal somebody else's version to blend with, as I did in #911 and #912.

Demonstration:
Ingredients: #914 only.
1) Add a duplicate layer and completely desaturate it, turn it into a grayscale.
2) Add as a layer mask the underlying blue channel. Now the desaturation is most pronounced in the front of the mountain.
3) Blend If to exclude anything that's dark in the blue channel on the Underlying Layer. This restores the background and limits the desaturating to the foreground.
4) Reduce opacity of the desaturating layer to taste. I chose 50%.
5) Create a composite layer of what we now have.
6) Auto Tone to establish a proper white point. Note also the desirable effect in the foreground.
7) Fade Auto Tone>Lighten 100%. Otherwise the trees get too dark and colorless.
8) Check to see if the Auto Tone layer works better in Luminosity mode so that the underlying blue doesn't change hue. In my opinion, the change to Luminosity mode does improve things, so I don't cancel this change.

The result is #935. Having now seen it and compared it to the other five par versions, I'd say that this foreground is my least favorite of the six, but I think I like the background best of all. So if the foreground were warmed up it might be my choice for best overall--Bill Theis, a Californian, has already posted that it's already his favorite among the pars. 

Another comment about this version appears at #927.

915 (97L 0a 2b) Chosen for the mysterious par. John Gillespie's original notes state:

The main task as I see it is to bring out more definition in the mountain without revealing too much of the banding and noise in the sky.

As averaging is a known strategy for reducing such artefacts I made a number of different versions using the Big Hammer, blending and also a false profile of 3.0 followed by a screen through an inverted luminosity mask. I also made a version with a very high contrast mountain, effectively turning the foreground into blackness.
I also made a very smooth version by adding noise then blurring.

I then blended these together with some additional darkening to the sky, then the MMM/CB and some sharpening to the foreground and the mountain peak.

As I noted yesterday, Bigger Hammer is not good for this particular image due to the lack of any definition in the snow. Also, although the background is interesting, the foreground suffered from the impact of some of the moves to alter the background. If John had done separate versions instead, there would have been no need for a false profile, he could have just multiplied the sky/mountain into itself.

Agreeing, afterward John writes:
I was keen to avoid too much banding and haloing (which I can still detect in the par) but the cost of this is less definition in the mountain in my version. Plus as usual the saturation is too low. 

916 (97L 0a 2b) The Mysterious Par itself. I don't think it's all that successful. Whatever you may think of the background mountains, the foreground is not as good as the "real" par. No surprise, IMHO: this particular image requires more experience than most of the others because of the difficulty in bringing out the mountain. People experienced enough to do that usually won't make big mistakes in something as simple as this foreground.

917 (91L 0a (3)b) Weak color, no highlight.

918 (100L 0a 0b)  Chosen for the par version. Excellent foreground; this is the lightest of the par parents.

919 (95L 0a (5)b) Gerald Bakker has already expressed his disappointment with this one but I think it simply reinforces the fact that two unrelated photos are forced to inhabit the same file. Gerald's foreground is arguably as good as the par's, in fact I prefer a 50-50 blend with the par to either parent, as far as the foreground is concerned.

The background, of course, is a very different story. See #923 for what happens when it's replaced.

920 (96L (1)a 0b)  In the smoke area at the front of the mountain this version consistently measures (5)a(7)b. Look how green/cyan it is next to 919.

921 (100L (1)a 1b)  Bill Theis used a combination of methods to successfully build the mountain. Overall, it is the least colorful of any of the top versions in both foreground and background. The overall luminosity, however, is fine. 

922 (94L (2)a (3)b)  John Castronovo: 

Although I liked what I saw on my screen when I did it because I thought it was reasonably accurate, looking at it now I don't know why I didn't blend the red channel into blue for a more spectacular backdrop. I just wasn't looking for that obvious blend option although it was slapping me in the face the whole time. More contrast and brighter wouldn't hurt either now that I look at it with fresh eyes.

To see what went wrong here call up #901, which is basically the same color.

923 (98L 1a 0b)  Chosen for the par version.This person worked on not two but four separate versions before blending them. He wound up with a synthesis between those with a mysterious look and those with a relatively sharply featured mountain. Much of his work came outside of Photoshop. His process notes state,

I have recovered detail in the top half but deliberately left the clouds hazy/wispy as I thought that would be more natural than too much detail in them.

Some amount of detail has been recovered in the mountain but again I have left it relatively hazy as attempts at detail extraction led to a blockish look.

The sky is a lighter blue, just a personal preference. The lower more solid clouds have been neutralized but the wispier ones have some blue in them. The darker areas of the mountain have had almost all -a removed thought there may be some small spots here and there.

I am not entirely happy with the colour separation of the lower half but after many attempts settled on what I have.

I agree with the above. What we have here is a background that's better than the par coupled with a foreground that's much worse. That makes it the counterpart of #919, which marries a poor background to a foreground that's very competitive with the par. Why not blend them?

Demonstration #1:
Ingredients: #919 and #923
1) Start with #919 and add a new layer.
2) To it, apply #923 in Normal mode.
3) The best Blend If will be found on the Underlying (#919) Layer. Exclude everything that is not light in the blue channel.
4) Reduce opacity to taste. I left it at 100%.  The result is posted as #936.

Knowing the taste of the father of #919, he would not have left that opacity at 100%. Gerald Bakker has declared himself an admirer of the misty-looking #926, and  comments on previous studies show that he consistently prefers more conservative treatments than I do. If I knew that he were the ultimate judge, I would do this blend slightly differently:

Demonstration #2:
Ingredients: #919 and #923
1) Start with #919 and add a new layer.
2) To it, apply #923 in Normal mode.
3) Convert the file to LAB, unflattened.
4) Add as a layer mask, the B channel of the Merged layer, Inverted. This favors the #923 layer where blue and the #919 layer elsewhere, but these are subtler blends than a strict Blend If. The result in the mountain is a hazier, vaguer shape than in #923/#925, yet much more prominent than #919. Meanwhile, the foreground gets a slight addition of the golden tone of #925, which seems helpful.

924 (87L 1a (3)b)  Very flat because of the heavy highlight. See the demonstration at #902 for how important that failing is.

925 (99L 1a 1b) James Gray has commented on this one and I can only add that a black sky and brilliantly white snow do not a convincing combination make. However, as an auxiliary file it's a worthy competitor to my #931. Anyone who feels that their own mountain is too weak can blend this one in with various methods.

926 (97L 1a (1)b) Chosen for the mysterious par, and our most commented-upon version, largely because of the interesting haze pattern in front of the mountain.
 
Jim Sanderson:
I found the Shasta image more difficult than the Pagoda picture from last week.  The Pagoda shot I thought had a better starting point to work from and I came up with a better result.  (Mine was 833)  I'm not as happy with the outcome of Shasta exercise when compared to the PAR version which's saturation I find more to my liking.  Not only the harsh mid day sun but the really smoky conditions had me trying various techniques without a clear preconceived plan of attack.  As a result, my layer stack in Photoshop was a mishmash of adjustment layers and masks.  Too many.  Now I see that I could have bumped up the saturation a bit more also.

Gerald Bakker replies:
The par is more saturated indeed, but I think this only works well for the foreground. 
Well, why not... I downloaded both 926 (Jim's) and 930 (par), to see if I could make a blend for an even better version. 

926: excellent background, maybe a bit pale
930: excellent foreground, background too much color, especially the sky
Start with 930 and add two copies of 926 on top. The first in Luminosity, the second in Color. 
On the Lum. layer, add a layer mask that is black for roughly the bottom half, white for the top half (best use a gradient to make the separation). For a darker sky, reduce opacity to taste (I chose 75%).
On the Color layer, add a layer mask en apply the Blue channel to it (Background or merged, doesn't matter much).
The result is a version that has most of 926's background and most of 930's foreground. I think it's better than both. 

Jim has responded that he agrees.

John Gillespie adds:
But getting rid of the smoky effect in its entirety in some ways reduces the interest of the image.
The Dictionary of Obscure Sorrows defines "vemödalen" as "the frustration of photographing something amazing when thousands of identical photos already exist—the same sunset, the same waterfall" -  the smoke provides a chance to have a different take on a familiar image.  Blending some of 925 (which has the most extreme "smokiness") in luminosity mode using "blend if" into the par produces an interesting take along these lines.

927 (89L (1)a (1)b) Chosen for the par version. For another reminder of why excessive blueness in the sky is a bad idea, compare this one to the more heavy-handed #914.

928 (94L 0a 1b) Robert Baldassano:

I live in northern CA, so have been on the roads where you can see Shasta looming ahead of you . I have been there when it has been hazy, and when it has not been hazy, but never when there was smoke. I have never seen it like the very dark images submitted. So my attempt was to see it through the haze. Lookong at some other images, I can see perhaps I should have made the foreground a bit more yellow, and perhaps have a bit more contrast, so I stacked  1930 in Luminosity mode @ 54% and 923 in Color Mode @ 80% and I think that looks better than mine alone.

929 (97L 1a (2)b) Compared to our better ones this seems somewhat dull. The colors aren't wrong, but the foreground is heavy and desaturated, while the background is flat and bluish.

930 (99L 0a 0b) The par version.

931 The auxiliary version for adding detail to the sky, produced as described at #909.

932 Alternate par #1, produced as described at #909.

933 Alternate par #2, produced as described at #911.

934 Alternate par #3, produced as described at #912.

935 Alternate par #4, produced as described at #914.

936 Alternate par #5, produced as described at #923.



Robert Wheeler
 

“906 (98L 0a (1)b) These numbers suggest that things might be a little green. Do you see it?”

 

Oh my. The heroic student finds himself struggling to escape from the canyon of ignorance, climbing hand over hand with great effort up a thin but sturdy cable woven of color theory facts (tethered to a trove of Lab knowledge above), then feels an jolt and unexpectedly slides precariously downward. Quick, free up one hand to review the guidebook. Hmmm.  With the “a” (magenta to teal-greenish-bluish) channel at zero, there was partial success toward having a neutral point. Then having the “b” (yellow to blue) channel at negative one, it says there would be an extra hint of blue. Hard to see “green” there, and hard to understand why the numbers “suggest” green. Maybe I need the cataract removed this year instead of next year. Wind picks up, cable begins to sway back and forth.

 

909 (94L (2)a 2b) … Given … the greenness of the measured highlight…”. Brace against this outcrop on the canyon wall. Teal negative “a” plus yellow positive “b” makes green. Grip tightens on sensible numbers here. What happened higher up the wall?

 

Patience grasshopper. A bit of extra blue added to the yellowish foreground could make “things” (other than the white point) look slightly green. Conundrum reduced. Grit teeth, grab with both hands, and resume the climb. Undaunted, the student continues to learn.

 

Robert Wheeler, perhaps after watching too many action-movie clips.


Gerald Bakker
 

On Thu, Apr 1, 2021 at 11:45 PM, Dan Margulis wrote:
909 (94L (2)a 2b) Chosen for the mysterious par. A pleasant misty look but it would be better with a white highlight. Given the choice to have a relatively gray sky and the greenness of the measured highlight, it might be worth trying a warming filter in the light areas, to get somewhat of the effect found in #903.
 
This person's High Andes entry is already in, and he commented as follows in his process notes:
 
I am learning so much by these comparisons and I realize that I still have big problems. For example, in the previous image (Shasta) I have not been able to understand how others have been able to add so much to the structure of the mountain without creating artifacts in the sky.
 
Like this:
I executed the procedure that follows, and find the result (uploaded as par #932) by no means "without artifacts in the sky". On the contrary, I would say. I would never accept this for myself, let alone show it to others. (Not sure how an average paying client would react to this result...)

My feeling is (but I may be wrong) that a well-defined artifact-free realistic-looking mountain plus sky cannot be attained without a lot of manual masking, painting etc. I tried what Dan describes about his own version 901: apply a number of different techniques, each bringing some detail into the background, thus hoping to keep the artifacts limited. Hmm, not quite successful. Working from what I got however, I was able to create a reasonable result with a background similar to #926 except for the smoky effect. But it took a lot of work, and it reminded me of what Jim Sanderson wrote about his version "... my layer stack in Photoshop was a mishmash of adjustment layers and masks." Same here.

Of all the pars, I prefer 930 and 936, except that I find both too heavy in the sky.
--
Gerald Bakker
https://geraldbakker.nl


Dan Margulis
 



On Apr 1, 2021, at 7:26 PM, Robert Wheeler <bwheeler350@...> wrote:

“906 (98L 0a (1)b) These numbers suggest that things might be a little green. Do you see it?”

 

Oh my. The heroic student finds himself struggling to escape from the canyon of ignorance, climbing hand over hand with great effort up a thin but sturdy cable woven of color theory facts (tethered to a trove of Lab knowledge above), then feels an jolt and unexpectedly slides precariously downward. Quick, free up one hand to review the guidebook. Hmmm.  With the “a” (magenta to teal-greenish-bluish) channel at zero, there was partial success toward having a neutral point. Then having the “b” (yellow to blue) channel at negative one, it says there would be an extra hint of blue.

My apologies, this was a typo when I transcribed from my spreadsheet. It’s 98L (1)a 0b. A similar reading, 96L (1)a 0b, is found in #920. These do suggest that things might be a little green. A suggestion is not the same as a guarantee. It is prudent, however, to investigate further because if this snow is anything other than 0a0b, greenness is worse than any warm color.

The next entry, #907, is further away from 0a0b than #906 is (97L 2a (1)b) but its cast is purple, close to the complementary of green. If we compare color only, by putting #907 on top of either #906 or #920 in Color mode 100% opacity, then I hold my nose and say that I prefer the new version in both cases. If we instead have the #907 layer at 50% opacity, Color mode, then not just me but everyone else in the world will say that both #906 and #920 have been improved. That happens to be the result that these numbers predict.

We are considerably more sensitive to color variation in light colors than dark ones, so it pays to be careful about highlight values. Again, sorry for the typo.

Dan


Dan Margulis
 



On Apr 2, 2021, at 7:02 AM, Gerald Bakker <gc.bakker@...> wrote:

I executed the procedure that follows, and find the result (uploaded as par #932) by no means "without artifacts in the sky". On the contrary, I would say. I would never accept this for myself, let alone show it to others. (Not sure how an average paying client would react to this result...)

Considering how many of us didn’t knock out the lens spot at upper right one would think that we are not seriously bothered by imperfections in the sky. #932 was not represented as a finished version. The person would have to decide whether to reduce the impact of the darkening layer, and also whether to change its hue and/or saturation. Presumably after doing this he could take the minute or so needed to erase such artifacts as offended him. It can be noted that in the mountain itself (not the sky) the artifacting is worse in the original (#909) than it is in #932.

If I were sending this out to a client I’d touch up the artifacts, but even if #932 went out as is I’d say a paying client would be a hundred times more likely to say the background was unrealistically dark than to complain about the artifacts.

Of all the pars, I prefer 930 and 936, except that I find both too heavy in the sky. 

As predicted in my comments to #923, from which #936 was derived. When I created the layer that established the heavy sky, I left it at 100% opacity, adding,

Knowing the taste of the father of #919, he would not have left that opacity at 100%. Gerald Bakker has declared himself an admirer of the misty-looking #926, and  comments on previous studies show that he consistently prefers more conservative treatments than I do. If I knew that he were the ultimate judge, I would do this blend slightly differently:

And I then proceeded to show a different method that would cater to your views perhaps better than reducing the layer opacity to 50% or whatever you would have chosen.

Dan


Gerald Bakker
 

On Fri, Apr 2, 2021 at 02:30 PM, Dan Margulis wrote:
As predicted in my comments to #923, from which #936 was derived. When I created the layer that established the heavy sky, I left it at 100% opacity, adding,
 
 
Knowing the taste of the father of #919, he would not have left that opacity at 100%. Gerald Bakker has declared himself an admirer of the misty-looking #926, and  comments on previous studies show that he consistently prefers more conservative treatments than I do. If I knew that he were the ultimate judge, I would do this blend slightly differently:
And I then proceeded to show a different method that would cater to your views perhaps better than reducing the layer opacity to 50% or whatever you would have chosen.
Agree, except for one thing (sorry, I'm nitpicking now). Reducing opacity of that layer lightens both sky and mountain. I'd rather prefer to lighten the sky only. But that's easily accomplished by a curves and/or HSL layer with a quickly painted mask over the blue areas of the sky.
--
Gerald Bakker
https://geraldbakker.nl


Harvey Nagai
 

The following Mondays of these case studies are always interesting, sometimes in a not very fun
way.  What seemed beyond perfection Sunday evening loses its lustre after seeing what other
people have done different and/or better, but it has always been a learning experience.

And so it was with my entry: #918.

This week my first thought was "Who the bleep sharpened my pic to death?".  Checking my notes
to Dan and the file history, it was me.  D'oh.

I was supposed to remember to NOT do it, so as to not draw more attention to the foreground. 
(The implications of two counts of not remembering may be even more disturbing.)

Beyond that I seemed to march to a different drummer out to left field in several other ways.

My interpretation of Dan's description that is was a "great day" with "bright, sunny weather"
seems to be "brasher" than average.  Maybe it's because I don't like to wear sunglasses, that's
the way I would have seen it (and my apologies to anyone who felt the urge to wear shades to
view my pic.)

I'm a little puzzled that more people (including Dan) didn't seem bothered by having a big area
of colorless denseness smack in the center of the image.  It may be realistic to show the trees
to be that shadowy, given the angle of sunlight, but I found it commanded too much attention
just by being obviously different and in a place where most people would look first.

My solution of lightening and coloring the trees to unrealistic levels and hues was less than
elegant in execution, but I think those who addressed the issue made it easier to notice the
mountain.

Processing the foreground and background separately was a natural decision for me, that Dan said
few people did so struck me as odd.  But maybe I abhor halos more than average, so my first test
was to see what the foreground would do if I simply slapped an extremely steep curve on the
background.  It went nearly black, which seemed to me to be a perfect mask.

I did get some pretty bad pixellation by doing that to the default image jpeg, but much less so
with a 16-bits version of it from ACR.  Only moderate noise reduction (with the ACR filter) was
needed at the end.

The overall strategy was to come up with a picture postcard correction, then blend back in the
default image or some other conservative version to what I thought Dan might have wished he had
seen, knowing that the view was going to be less than crystal clear.  Maybe I didn't blend out
enough of the foreground and a bit too much of the mountain.

If I could pick one partner to blend with mine, I'd pick #904 for its excellently realistic
conifer greens, put mine on top and mask with the blue channel, adjust the mask density to taste,
tweak a bit of green into the blue sky.  That would dumb down the brassiness in the low foreground
and give the trees much better color.

My thanks to Dan for such informative commentary, as to those who participated.


Robert Wheeler
 

Ah, mundane typo more than extra obscure way of making me look at the whole picture. Appreciate the clarification and the additional teaching points still emerging in the thread. Thanks.
Robert Wheeler


Pgallivan@axion.com
 

😎

Pat Gallivan 
———————
fmg@...
   303-547-5910

On Apr 1, 2021, at 5:26 PM, Robert Wheeler <bwheeler350@...> wrote:



“906 (98L 0a (1)b) These numbers suggest that things might be a little green. Do you see it?”

 

Oh my. The heroic student finds himself struggling to escape from the canyon of ignorance, climbing hand over hand with great effort up a thin but sturdy cable woven of color theory facts (tethered to a trove of Lab knowledge above), then feels an jolt and unexpectedly slides precariously downward. Quick, free up one hand to review the guidebook. Hmmm.  With the “a” (magenta to teal-greenish-bluish) channel at zero, there was partial success toward having a neutral point. Then having the “b” (yellow to blue) channel at negative one, it says there would be an extra hint of blue. Hard to see “green” there, and hard to understand why the numbers “suggest” green. Maybe I need the cataract removed this year instead of next year. Wind picks up, cable begins to sway back and forth.

 

909 (94L (2)a 2b) … Given … the greenness of the measured highlight…”. Brace against this outcrop on the canyon wall. Teal negative “a” plus yellow positive “b” makes green. Grip tightens on sensible numbers here. What happened higher up the wall?

 

Patience grasshopper. A bit of extra blue added to the yellowish foreground could make “things” (other than the white point) look slightly green. Conundrum reduced. Grit teeth, grab with both hands, and resume the climb. Undaunted, the student continues to learn.

 

Robert Wheeler, perhaps after watching too many action-movie clips.


David Remington
 
Edited

Dan,
 
Thanks for the review of my version 914 and all the other demonstrations.
 
I agree the sky is overpowering and unnaturally blue. I'm tracing this back to applying dehaze in Camera RAW. I applied it through a gradient to the top of the image. I could see that it pumped up the sky but did not think it would be an issue. I should have used an adjustment brush on the mountain alone.
 
I also agree with Gerald that many of the images and examples have significant artifacts that would not be acceptable in a final image. As I see it though, and as you responded, these are not final images but rather examples of how to apply a technique. How or whether  to use them and refinement is up to individual user.
 
I am learning some new techniques, experimenting, and in particular with the more recent examples, taking a much more aggressive approach than I would in my regular work. This is a good place to see what works and what does not work.
 
David