Date   

Re: A Toast to Greece: Results

Rex Waygood
 

On Tue, Jul 14, 2020 at 10:49 PM, <jwlimages@...> wrote:
One thing I'm not so happy with now - the greenery behind the people on the left side is too bright & saturated, I should have addressed that.
I think one of the strengths of the exercise is that you get to see the image with the eyes of others. It helps with seeing how your own version could have been improved.
Rex


Re: A Toast to Greece: Results

Rex Waygood
 

The floor also seemed too bright, and the terra cotta-looking bowl in a prominent intersection just jumped out at me, so I retouched it out (what is that, anyway?). I see three other people also got rid of it.

Thereby eliminating memories of the family dog?  :-) :-)

Rex


Re: A Toast to Greece: Results

Robert Wheeler
 

Mine is 925. I experimented briefly with gradients to mask color transition areas only to quickly prove to myself that I need to learn a lot more about gradient techniques before I can actually use them. Interesting to find that others used them productively, so I'll move the topic higher up in my learning plan.
Robert Wheeler


Re: A Toast to Greece: Results

Thomas Hurd,MD
 

I took each image and loaded it as a stack. I compared each in normal, color and luminosity mode to the par image at the bottom.

I found the par image to still have too much blue cast, and perhaps that is why I found many more entries to improve or be equal to par in color blend mode. Just a few improved, or did not worsen, the par image in luminosity blend mode.

Color mode equal or better:
901 (at 50% opacity)
906
907 improved less yellow on the right
908 slightly cooler/bluer
913
915 (mine, I will say more)
916*
918
919*
921*
924*
925
927
930* very slightly bluer 

Luminosity mode equal or better
902 slightly darkens but a good blend
914
916* right side a little darkish
919*
921*
930*

My color correction was hampered by the fact there were mixed casts. I couldn’t find any good channel masks to help, so after I neutralized my highlight and shadow, I had to use gradient masks to get essentially four different pieces to the neutral not bad color puzzle. I went around once to fix the table cloth and a second time with the faces, decreasing blue to get enough yellow to be at least equal to magenta. It sounds like John Lund and Gerald Bakker used gradients in a somewhat similar fashion.

I did some work on the noise, but as I scrolled through all the entries, the noise didn’t affect my appreciation of the image one way or the other.

Somewhere in my workflow I did blend the green channel into the red with a layer luminosity blend, but it doesn’t look like it as I look at my downloaded image. Repeating that step makes my entry better looking I think. I tried the same technique on 927 and 930 with better results I think.

Overall my favorites were 930, 925 and 907. I liked my own 915 best for the hues but I was missing some contrast and weight balance.

To John’s diversity comment, its interesting that 6 images were mentioned favorably, but only once so far, and 6 were mentioned multiple times. 


Tom Hurd



On Jul 14, 2020, at 5:49 PM, jwlimages via groups.io <jwlimages@...> wrote:

Again, I found the range of efforts interesting and surprisingly large - a reflection of the diversity in this group?

 

I thought this image was in some ways pretty straightforward. It's underexposed of course, probably due to the photographer shooting directly into the two spotlights at rear. I was surprised to see that while pretty much everyone brightened the overall scene, some people enhanced the "candlelight" effect, leaving the people in the shot relatively dark. A couple folks went to work on the spotlights themselves, darkening them to the point of minimizing them, and one person completely retouched them out.

 

The relatively cool light of the on-camera flash dominates the left half/foreground, while the folks at the rear table are painted with the orange-yellow of incandescent lights. It looks like there were several different kinds of responses here as well, from cooling or warming color overall to desaturating, to "re-lighting" with a spotlight effect.

 

Anyway, for my entry (#930): I started with the premise that the most important task is a pleasing rendition of the people. I re-processed the jPeg through Lightroom, increasing exposure & contrast, lifting shadows plus adding some Vibrance. Then I used graduated filters to darken the left half of the shot while adding exposure & decreasing color temp from the right third of the image. The resulting was surprisingly good - I almost submitted it. But hey, this is an ACT exercise, right? and I want to learn to use PPW tools better, so I continued with the Lightroom re-processed file. It needed some noise reduction to start with, then after global "brightening" & contrast boost, complimentary color corrections to the two halves of the shot. I then used MMM+CB (greatly dialed-back intensity), a narrow strip of H-K layer (again lowered opacity) for local contrast in the people, and with a couple final tweaks got to pretty clean color & pleasing contrast (I think).

 

A couple compositional elements still bugged me - the leafy canopy of the tree seemed way too prominent lit up by the flash, and distracted from the people. So I darkened & desaturated it. The floor also seemed too bright, and the terra cotta-looking bowl in a prominent intersection just jumped out at me, so I retouched it out (what is that, anyway?). I see three other people also got rid of it.

 

One thing I'm not so happy with now - the greenery behind the people on the left side is too bright & saturated, I should have addressed that.

 

John Lund

www.jwlimages.com



Re: A Toast to Greece: Results

jwlimages@...
 

Again, I found the range of efforts interesting and surprisingly large - a reflection of the diversity in this group?

 

I thought this image was in some ways pretty straightforward. It's underexposed of course, probably due to the photographer shooting directly into the two spotlights at rear. I was surprised to see that while pretty much everyone brightened the overall scene, some people enhanced the "candlelight" effect, leaving the people in the shot relatively dark. A couple folks went to work on the spotlights themselves, darkening them to the point of minimizing them, and one person completely retouched them out.

 

The relatively cool light of the on-camera flash dominates the left half/foreground, while the folks at the rear table are painted with the orange-yellow of incandescent lights. It looks like there were several different kinds of responses here as well, from cooling or warming color overall to desaturating, to "re-lighting" with a spotlight effect.

 

Anyway, for my entry (#930): I started with the premise that the most important task is a pleasing rendition of the people. I re-processed the jPeg through Lightroom, increasing exposure & contrast, lifting shadows plus adding some Vibrance. Then I used graduated filters to darken the left half of the shot while adding exposure & decreasing color temp from the right third of the image. The resulting was surprisingly good - I almost submitted it. But hey, this is an ACT exercise, right? and I want to learn to use PPW tools better, so I continued with the Lightroom re-processed file. It needed some noise reduction to start with, then after global "brightening" & contrast boost, complimentary color corrections to the two halves of the shot. I then used MMM+CB (greatly dialed-back intensity), a narrow strip of H-K layer (again lowered opacity) for local contrast in the people, and with a couple final tweaks got to pretty clean color & pleasing contrast (I think).

 

A couple compositional elements still bugged me - the leafy canopy of the tree seemed way too prominent lit up by the flash, and distracted from the people. So I darkened & desaturated it. The floor also seemed too bright, and the terra cotta-looking bowl in a prominent intersection just jumped out at me, so I retouched it out (what is that, anyway?). I see three other people also got rid of it.

 

One thing I'm not so happy with now - the greenery behind the people on the left side is too bright & saturated, I should have addressed that.

 

John Lund

www.jwlimages.com


Re: A Toast to Greece: Results

Gerald Bakker
 

I found this one of the most difficult exercises so far. The main focus of the image is the table and the persons. The large majority of these elements is white. As a consequence, PPW doesn't work very well. Color Boost would emphasize the leaves, the flags, the floor tiles and the blue shirt on the left - none of these makes the image particularly better in my opinion.

Also there was something odd with the color. The table cloth is supposed to be white or close to white, but when I made it white, the faces got too pinkish (A equal or even larger than B). So I decided to leave a bit of yellow in the whites. (My version is #907). Still, not an easy thing to find the best hue overall.

Then, how light do you want the image to be? I prefer the lighter versions, after all white is white. Being at the scene, we would perceive the cloth and white shirts as white and not some middle gray. But the danger is that detail gets lost. As an example, #908 has good, strong whites but the detailing is subpar. #904 on the other hand is on the dark side, but the folds in the cloth look very good.
#919 does both very well (definitely one of my favorites).

To counter the disbalance in color and luminosity, I applied a few gradient masks. Effective and invisible. No need to make complicated selections or apply some magic in LAB.

Noise wasn't such a problem. What was there was not very obtrusive, and could easily be reduced.
--
Gerald Bakker
http://geraldbakker.nl


Re: A Toast to Greece: Results

Hector Davila
 

Mine is #901

I have a different view of what others call "noise".

I have made no attempt to remove the noise or grain because as I see it...

when a person takes a picture with their camera,

 the picture.. comes with noise!

I like to keep the picture as natural as possible
because to me.. the people are most important.

I want to be able to capture the personalty of a person,
their soul...and those features are mixed in with
the details of their features and the noise/grain.

I believe, if you remove the noise, you remove the soul.

Just removing one single catch light from a person's eyes will remove their soul completely.

When the photo is printed and passed around ...everyone only looks at the photo of themselves only.

That's a lot of people to please.

So I make them as bright and colorful as possible. Bring them ...back to life.

Hector Davila


Re: A Toast to Greece: Results

Robert Wheeler
 

The family reunion image posed multiple challenges that helped me learn more in several areas.

 

Noise. EXIF shows ISO 2000. Sky shows diffuse light colored grains consistent with luminosity noise, and the faces of people at the far table have irregularities consistent with noise. When I attempted corrections initially, the noisy faces all became more distorted and blotchy. Lesson learned: fix the noise first. Early attempts at noise removal led to faces being excessively smooth with diminished detail that was only partly corrected by applying the green channel to the red. Tried Topaz AI method with low settings without great improvement. In the end, I made a duplicate background layer and applied the camera raw filter with low luminosity noise reduction that seemed to work well enough.

 

In the submissions, 922 has a pattern of white dots in the dark sky consistent with stars. I don’t see this pattern in the original, making me suspect replacement with night sky from another image, but would be fascinated if some technique removed noise while leaving stars. 924 has a sky that looks like snow. Family members would probably ignore even the noisiest sky, but the various images with blotchy or distorted faces seem problematic.

 

Color. The people at the back table have very warm skin tones (reds, yellows) consistent with tungsten and candle lighting wile the people at the near table have cool skin tones consistent with other light, possibly flash. All could benefit from better color. I found that the skin desaturation action helped moderate the impact of MMM+CB, but did not effectively deal with the difference in lighting. I tired quite a few variations of channel and luminosity masking without finding a good combination for this. Desaturating and moving everything to the blue end of the scale did allow more uniformity but looked uninviting and depressing. I might have tried applying color look up tables, but these are beyond my current skill set.

 

Pondering about LUTs led me to consider a solution involving dishonesty about the color right from the start. So, after noise reduction, I used the camera raw filter “calibration” section to falsify the reds as being +15 up the color scale from “reality.” In the calibration section, I also desaturated the reds and blues both to negative 15. This made all the skin colors much closer together. Experimenting showed I could now skip the skin desaturation action. Running MMM+CB with a rough selection of faces, tables, and flags, and using the endpoint layer on auto produced quite strong colors (but skin fairly consistent among the group). Turning off the MMM saturation layer and MMM color layer helped somewhat. On return to sRGB, two additional steps helped. First, a hue-saturation layer lightening yellows to +20 and reds to +10 and boosting master saturation to +20. Then a final adjustment applying the camera raw filter to move color balance negative 10 (mild reduction of yellows that were still too strong) but tempered with blend-if to exclude reduction in the light blues (allowing the closer tablecloths and shirts to retain some of the inviting warmth and make color across the group a bit  more even).

 

Two dark faces. The seated man with an iPhone and the elderly-appearing woman at the apex of the tables both started with darker faces than most. I was not able to find global masking that would help them while not hurting others. I finally gave up and added a curves layer with points added to moderately lighten the RGB curve, slightly lighten the red curve, and slightly darken the green curve. Black mask, then painted just the two faces with low opacity white to selected parts of face/neck/hair with some improvement in color and impression of detail. Many of the entries had reasonable results with these faces, so I’ll be watching for clues about more efficient approaches.

 

Overall. I was surprised by the wide range of submitted results. The most favorable ones to my (non-expert) eyes are the 931 PAR version along with 908, 911, 921, 925, 927, and 930. Admittedly, one of those is mine, which I will identify later after allowing time for others to submit their comments first.

Robert Wheeler


Re: A Toast to Greece: Results

John Gillespie
 

Fascinating to see so many different interpretations of the same image.
Broadly speaking there seem to be two main approaches - either making the people as bright as possible or attempting to retain (or enhance) the nighttime atmosphere.
The par version is a very good rendition of the first approach, and this most fits the brief I think. 
However the second approach possibly makes for a more interesting image. 
In the second camp I would put 904,  912, 920, 921, 925 and 926, with 926 getting closest to the brief. 
904 and 912 are the most "dramatic" and perhaps closest to how we would have experienced such a scene "in real life".
If this was a different type of gathering (one in which the top table guests are  more important than the others) then 920 could be a strong contender.
An average version of all these images is I think excellent and a strong contender against the current par version.
911 is a good compromise between the two approaches. The masking is obvious when seen along side the others but when viewed standalone this is not a problem.

Of the brighter versions 906,914 and 930 are to my mind the best examples.They have managed to reduce the blue cast from the flash without making the scene too orange, but still retained the warm summer evening feeling. The faces have fairly even lighting and the overall contrast is good. 









moderated Case Study: Seated in the Grass

Dan Margulis
 

Three more case studies to go, so While discussion continues of the Toast to Greece shot, I’ve posted our final portrait, another selection from the MIT study. Therefore, as with the other MIT exercises, unsharp masking is not permitted.

Our final three case studies try to close the circle that we began to draw with the outset of the series. This one isn't as difficult as Veiled Bride or Niagara Spray, nor does it have the lighting issues of A Toast to Greece, but it should be challenging enough.

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

*This is one of 5,000 images taken from a large MIT study. for which permission is given for educational use. The sponsors solicited contributions of images that were supposed to represent the range that a professional retoucher might receive. They then hired five intermediate retouchers to correct each. The 25,000 resulting files were made public, as were the .dng files they started with. These efforts will be posted along with those of our group.

*In the study, no instructions were given as to what the client wanted.We assume it is some kind of family keepsake.

*You can use whatever methods you like to improve the picture EXCEPT AN UNSHARP MASK filter or similar, because the original retouchers in the MIT study were not allowed to.

*Please keep clear records of what you did for discussion. List members find these very valuable.

*In the Photos section, Case Study: Seated in the Grass, 
I have uploaded a version opened with Camera Raw defaults, and another where the settings were much flatter. You may use either, or fetch the .dng as below.

*groups.io does not allow .dng format in the Photos section. If you want the .dng, you must download a zipped file from the Files section. NOTE: the zipped file contains the two default images as well, you don't need to download them separately. Filename=Seated-in-Grass_case-study_source.zip

*The designated size of this exercise is 2000 x 3008 pixels. If you use the .dng image be sure to open into the correct size. Do not crop, rotate, or alter the sizing, and don't delete any objects, because any of these things will make it impossible to use your version as part of a par assembly. Also, we have had that Lightroom was not acquiring these .dngs in exactly the same aspect as other modules due to a lens correction. Therefore, we recommend that as soon as you acquire, you apply one of the default versions to it to make sure that all pixels line up.

*Your final file is to be sRGB with a proper tag. If you work in a different RGB you must Edit: Convert to Profile>sRGB before submitting the file.

*When finished, save in JPEG form, quality level 9. E-mail it to me, dmargulis (at) aol.com, with a brief explanation of how you produced it, including what file you used for your original. DO NOT POST IMAGES TO THE LIST.

*Remember that some e-mail clients automatically downsize image attachments. Make sure you’re sending it to me at the original size.

*Entries close Monday morning, 20 July, at 06:00 Eastern/1100Z/12:00 ora italiana.

*Rather than confirm every entrant I've received, I will periodically post the initials of everyone whose file I have.

*As soon as convenient after the deadline, I'll post all the entrants in a random order. Names will not be revealed except for those entrants who have identified themselves publicly. 

*A discussion will follow within a few days after posting the final files. 



A Toast to Greece: Results

Dan Margulis
 

I’ve posted the results of our latest night scene.

Reviewing: This is the welcome dinner for a very emotional reunion: a Greek family scattered all over the planet by World War II getting back together in the mother country seventy years later. The shot is taken outdoors at night, with lighting that varies from one side to the other, making the exercise quite difficult. I’ve given it to one class; there are a couple of their results posted, not among the best you’ll see.

We have 30 entrants. When a person submitted two or versions, I chose the one I thought was better. Most people also submitted a list of their steps, thanks very much. I haven’t read these, because I’d rather get a sense of who was successful and who wasn’t before investigating why. 

The files don’t have people’s names on them, and were random-generator numbered from #901 to #930. As with past studies, we also have a “par” version, #931. To get it, I chose what I thought looked like the five best entrants, and averaged them, each one weighted 20%. This often creates a version that is superior to most if not all of its parents.

I’ll have some things to say about this assortment, but as usual I’d like to open it up to group discussion first. What do the successful versions have in common? Meanwhile, if you’d like to know how your own version stacked up, download the par version and compare the two directly. Do you think you got the same kind of quality? If not, I hope you’ll find further discussion useful.

The folder is in the group Photos section, named Case Study: A Toast to Greece

Because some of us would like a closer look at these, I also have zipped all 31 and uploaded a 46 mb file to our Files section,
Search for Toast-to-Greece_entries_071320.zip

I look forward to your comments.

Dan Margulis

P.S. The next case study is announced today, look for a separate post. The countdown continues, we have three case studies left.


Re: Monument Valley: comments on individual versions

James Gray
 

I am confessing that 815 is mine.
In case anyone is interested, I finally went back and looked at what I did that caused the local contrast that Dan did not care for.  Most of the local contrast that seemed to damage the color was not caused by the use of the ALCE plugin.  It was caused by an action I got from a friend that uses surface blur to exaggerate details.  I do not think that ALCE hurt it.  I think Dan might agree if he saw the version without the use of the action from my friend.  I also discovered that I used a curves layer after the MMM+CB step to mute colors.  It looked like that layer exaggerated the light halos.  I thought most of the entries were too saturated.  I really like the effects of ALCE.  It is true it can sometimes create violent effects, but with the use of masks and opacity reduction the effects can be subtle. 

James Gray

On Wed, Jul 8, 2020 at 5:51 AM Dan Margulis via groups.io <dmargulis=aol.com@groups.io> wrote:


814 The color is correct and the sky is nice, but the lack of variation in the foreground stands out. 

815 A complicated series of blends led up to what was apparently an overdose of Davide Barranca’s ALCE plugin which, being a form of sharpening, should probably not be permitted here but hasn’t specifically been banned. But it can give us an important reminder about sharpening practice.

In #814 we just saw an ultra-conservative counterpart to this violent version. Blending the two can be instructive but would generally be a matter of taste. Here’s one that isn’t a matter of taste IMHO: #815 suffers from the equivalent of light haloing, which damages some of the color. Generally light halos are worse than dark ones. Blend #814 in at 40%, Darken mode, and these halos are much improved. (Since #815 is basically a darker version than #814 except in the halos, this blend mode works well to target them.



Re: Seeking Clarification of Procedure for Averaging Images

Thomas Hurd,MD
 

Rick,

Your  hunch is correct.

  • Just took a 5 layer stack in VARIANCE mode. 
  • Converted to LAB.
  • Took the L channel from the variance mode stack and placed it’s mask on 510 over 511 blended in luminosity.
  • The yellow from 511 came through with the mask much more than luminosity with no mask.
So, in this particular instance, looks like the bright values in the Variance stack L channel are the luminance of the yellow flowers.


Tom






On Jul 9, 2020, at 7:29 PM, Rick Gordon <lists@...> wrote:

I haven't experimented yet, but I have a hunch that masks made from the L, and maybe even the a* and b* of a standard deviation stack rendering, maybe auto toned or maybe not, might offer some intriguing possibilities, maybe somewhat analogous to what can be done with auto toned difference masks.

Even the a* and b* channels of a standard deviation stack rendering show the contours of objects in ways that I bet, used as masks, could lead to some interesting results.

I've got to check that out.

Rick Gordon

--------------------
On July 9, 2020 at 4:17:35 PM [-0700], Thomas Hurd,md Via Groups.io wrote in an email entitled "Re: [colortheory] Seeking Clarification of Procedure for Averaging Images":
I expected a bit more from standard deviation.
___________________________________________
RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________
WWW: http://www.shelterpub.com


moderated Re: Case Study: A Toast to Greece

Dan Margulis
 

A reminder that entries are due in this case study in 48 hours, at 06:00 eastern time Monday/1100Z/12:00 ora italiana.

As with the last time two entries have already come in in the wrong colorspace. Please make sure you are sending tagged sRGB. I have converted the incorrect ones.

I confirm receipt of entries from the following individuals:

GB
RB
HD
RG
HH
AM
DS
JT
RT
ReW
RoW
*indicates that a corrected version was submitted

Entries from the following were at an incorrect size/cropping and would have to be resubmitted:
PM

Dan


Re: Monument Valley: comments on individual versions

Jim Sanderson
 

Jim Sanderson here.  With regard to 808.  After looking at the other entries, it is a bit flat.  "It looked good at the time." or famous last words.  Anyway, I corrected the sky and and moved the end points in on a levels adjustment and it did look a lot better.  I did try the auto tone both before and after the curves adjustment and in both cases it turned the clouds cyan and added a green cast to the entire image.  The auto tone also flattened the image further.  In looking at the histogram, all channels moved to the right (got lighter) and away from the left?  Thanks very much for the instruction and comments.

Jim Sanderson 


-----Original Message-----
From: Dan Margulis via groups.io <dmargulis@...>
To: colortheory@groups.io
Sent: Wed, Jul 8, 2020 4:51 am
Subject: [colortheory] Monument Valley: comments on individual versions

Here are my comments on the individual versions of this exercise, with some discussion of the techniques used. I start off by saying there are at least a dozen excellent versions here so that my choices for which ones made par are quite arbitrary,

Dan Margulis

801 This is the averaged version of the five MIT retouchers. The sky came out well, but without an adequate means of creating variation in the foreground the result is boring. So is #826, which is the one I judged the best of the five.

802 This excellent version shows off the PPW advantage in this type of image; just compare it to #801. This person did some mild adjustments in ACR, but he then wheeled out H-K and MMM, being properly conservative with his color boosting. He also made a CMYK copy to steal a black channel for blending. As a reminder, that can be done automatically by option-clicking H-K in the panel. All these things, plus a luminosity blend with the green channel, created great definition in the butte.

Also, note that he could have created a whole lot more definition in the clouds at the top of the sky, but chose not to. I think that was the right decision. No point in creating a diversion from the main attraction.

803 Chosen for the “conservative par” version. Granted the lighter foreground, it makes sense for this version to have a more detailed and darker sky than #802 does. Robert Wheeler created a separate contrast version outside of Photoshop, which accounts for decent detailing, but did not use MMM. His full description is posted in the main thread.

804 Chosen for the par version. The coloring is extravagant but I’m not going to criticize it in view of what a cliché the image is. His steps are basically the same as #802, just done with a heavier hand, note how much yellower the near foreground is than the butte itself.

805 As #804 is to #802 this one is to #803: same basic concept but more intense coloring. This one could have benefitted from H-K to create some more neutral areas. Also, the person describes a complex series of six versions, each with their own problems and strong points, which he then blended to create his final. And with all that effort, taking the second or two to try out Auto Tone at the end would have made a nice difference.

806 Here’s a straightforward assignment of contrast without much color variation. In both this one and #805 the upper right blue sky is very different from what’s at lower left, which I find distracting.

807 I did this one in 2017, limiting myself to three minutes. As with all the MIT images in this category (deserts, canyons, etc.) PPW showed a huge advantage. Here, for example, my competition was #801. I hope you will agree that #807 is decisively better, which is how I rated it.

808 The individual did not state a workflow but the end result is a flat image. Also, the lightest clouds are blue, not white. Just try Auto Tone at the end and it makes a major improvement in both color and tonality.

809 Chosen for the “conservative par” version. Very pretty. Attractive sky due to whiteness of the clouds. Although the clouds are detailed he is careful to avoid giving them enough punch to compete with the foreground. He adopted a sensible strategy for this image. First, he made a version entirely in Capture One. Then he started over and went crazy with PPW. There was no harm in doing this because he had a much tamer version available for blending. In fact, there’s often a lot of good in looking at something overly colorful, it helps us understand what the danger areas are. After blending the two, he did some manual adjustments, such as painting in the yellow flowers.

810 Chosen for the par version. Gerald Bakker is not a big fan of some of the more spectacular renditions of this scene and was looking for something more conservative. His approach was basically PPW, but his first pass struck him as too contrasty in the butte, so he made a second pass with the intent to blend.

Now we turn to artistic intent because Gerald made a couple of decisions that others specifically reversed. First, he decided near the end of his process that his greens were unnaturally saturated, and he toned them down. This disagrees with some who said they made them more intense, thinking that it would complement the reds of the landscape. Second, he applied his MMM Finetuned action to the sky. It produced a very soft and interesting effect. Several people, however, tried to avoid this as being too distracting from the butte. I had this experience myself in my #824; I originally put in a sky similar to Gerald’s and then decided it had to be toned down.

Since this came right after #809, which is also somewhat conservative, a comparison seems inevitable. #809 avoids the complex sky, and makes the clouds very white. It also is slightly more of a golden color. The greens are lighter and more saturated, especially in the yellow flowers.

I therefore rate #809 better for color and would describe its look as warmer and friendlier. #810, however, looks more realistic while retaining interesting variation. So in choosing for the original par version I threw up my hands and took #810.

Not before, however, attempting a blend. It seems logical that if #809’s color is better then blending it into #810 in Color mode would make an improvement. It does not. These two versions are integrated enough that neither can accept either the color or luminosity of the other without damaging the concept.

The way to get the best of both: blend #809 into #810 in Darker Color mode. This preserves Gerald’s sky and subdued greenery, while adding nice accents to the butte. I think that would be my favorite entrant, if it existed.

811 Another very nice version, notable for its treatment of the yellow flowers in the greenery. It is hard to say how this was accomplished, since the person was working mainly in Photoline, not Photoshop. Like several others, he treated the sky completely different from the rest of the image. He also selected the butte itself for special attention, moving it toward a rust color and adding contrast. So as I understand it he did not use MMM or equivalent, but tried to get a similar effect via selections and maskings. To me, the result is a little busy, with the greenery, the yellowish foreground, the butte itself, and the very active sky all competing for attention. But who’s to say this isn’t right?

812 Chosen for the par version. The person writes, “Right now the image looks colorful and well saturated, it will be interesting to see if I still believe that after seeing the other entries.” I have to agree that it is. The impression is very powerful and it seems like the saturation is much more than #810. In fact, they are about the same. The big difference is the overall weight. That this one is heavier also makes it seem more powerful colorfully. It only gets away with this by strong doses of MMM and H-K, otherwise it would become a fireball like ##805, 813, 821 and others.

813 Extremely red, not helped by the fact that the background clouds are quite blue.

814 The color is correct and the sky is nice, but the lack of variation in the foreground stands out. 

815 A complicated series of blends led up to what was apparently an overdose of Davide Barranca’s ALCE plugin which, being a form of sharpening, should probably not be permitted here but hasn’t specifically been banned. But it can give us an important reminder about sharpening practice.

In #814 we just saw an ultra-conservative counterpart to this violent version. Blending the two can be instructive but would generally be a matter of taste. Here’s one that isn’t a matter of taste IMHO: #815 suffers from the equivalent of light haloing, which damages some of the color. Generally light halos are worse than dark ones. Blend #814 in at 40%, Darken mode, and these halos are much improved. (Since #815 is basically a darker version than #814 except in the halos, this blend mode works well to target them.

816 In principle this is an adequate conservative treatment with many of the features that characterize the best results. It ought, therefore, to be directly comparable to Gerald Bakker's #810, which also answers that description. Unfortunately, #810 is clearly better. A casual observer might say it is because #810’s sky is much more interesting without being a distraction.

The real reason, though, harks back to the village in Cinque Terre. Everybody will find a highlight in the clouds and has a choice of shadow areas, so everybody will have approximately the same overall tonal range (particularly those who take my advice and check with Auto Tone after “completing” their correction).

But when one object or area dominates the scene, as the village did in Cinque Terre and the butte does in this image, then the question is how much range is assigned to that. To find out, I measured a large relatively light rock near the bottom of the butte and a shadowy area at its top right. In #816 the range is 65L-12L, 53 points total range, whereas in #810 it’s 76L-9L for 67L total range.

This is why #810 is the better of the two versions. And the “conservative par” version, of which #810 is not a parent, has almost identical values: 76L-8L.

817 Chosen for the “conservative par” version. This is the yellowest of our serious contestants. The technique is straight PPW starting with the flat version, including H-K, and the Lesser Hammer action aimed specifically at the sky.  The person stresses, “I believe that this exercise shows and confirms (as if we needed confirmation) the importance of an accurate correction of color before any other steps. I have tried this one with several curves and the result varies greatly even with small variations in the color curve.”

818 A reasonable non-PPW version. Why no PPW? Because this guy is using Photoshop CS3, which is indeed a capable program.

819 The foreground is reasonable PPW, no H-K, but a luminosity blend from the blue to the red. The person was concentrating so much on the foreground, which came out well, that he left the darker clouds much too blue. Clouds do get bluer as they get darker, but not so much as that. 

To see how important this is, grab a copy of #817, which has nice white clouds, and paste it as a layer on top of #819. Add as a layer mask the blue channel from the bottom layer. As this mask is quite dark, basically it substitutes one sky for the other while not changing much in the foreground.

820 Flat and washed out compared to most others. The clouds were not used as a white point as they should have been. By placing a white point in the lightest clouds and then darkening the midtone overall the version becomes much more acceptable.

821 Chosen for the par version. John Lund writes, “ I  found this image a simpler challenge than even the Cinque Terre, much less the other ones, mainly because it’s a decent capture of a straightforward landscape with pretty simple lighting. So no need for heroic measures to extract something usable.” He states he is familiar with the colors of the area from trips to Utah, which has similar lanscaping, but was not intending realism. Instead, he used Lightroom to enhance the colors of everything; the oranges, the sky, and the greenery. He then moved into LAB and added MMM+CB, saying “I liked the overall ‘hyper-saturated’ look. Nevertheless, he came back the next day and added an H-K layer, Luminosity mode, to get more tonal contrast in the butte. I don’t understand why he wouldn’t have also used H-K to subdue some of the more neutral areas and make the overall color less oppressive. But, it is what it is. Unsurprisingly, John vindicated my earlier prediction when he announced in the main thread that he prefers the main par version to the “conservative par”, putting him at odds with others. But, as his private comments make clear, “The final may be a bit over-cooked, but that’s how I experience those wonderful Utah landscapes, so maybe I can get away with this under the rubric of ‘pleasing color.’”

822 This person describes the image as one trap after another and says he eventually decided that less is more. It’s a reasonable conservative effort.

823 Chosen for the “conservative par” version.  This very nice rendition has the rosiest butte, and is thus the counterpart of #817, the yellowest one. This guy did much of his contrast-enhancing work in CMYK. And why not? It’s not like there are any brilliant colors that are out of the CMYK gamut. And so he got better handing of shadows than many did by paying special attention to the black channel. He also added a golden tone to the highlights and used an unspecified channel mask to prevent it transitioning into the butte.

824 Chosen for the “conservative par” version.  It’s mine, and features H-K, MMM+CB, and the Darken Sky action. A couple of days after doing it I reopened it and decided that the effort I had put in trying to make the sky fascinating (including Bigger Hammer) had been counterproductive, so I drastically lightened the upper clouds and am glad I did.  Interesting to compare it with #807, the one I did in 2017. Looks like my sky was too heavy-handed there.

825 Interesting that this one comes right after the discussion of the sky in #824, because this person had the same idea, Bigger Hammer plus Darken Sky actions to make something interesting happen in the sky. It is similar in conception to #824 anyway, if I had to choose one I’d pick #824 because of superior contrast in the butte (see the discussion of #816 being compared to #810), however this #825 has slightly better color in the clouds and the greenery. So, if you put #824 on top of #825 in Luminosity mode, I prefer the result to either parent.

826 This was the best individual result from the five MIT retouchers. It can speak for itself.

827 An outlier, a version that isn’t similar to any other. It takes the philosophy of #814 that the sky should be lightened a bit farther. Here, the clouds are suppressed to much that the butte seems almost to be silhouetted. This person identifies the shot as being a winter scene based on what seems to be frost in the foreground. He describes an exceedingly complex workflow that I don’t fully comprehend. I think I underrated the unusual result, because it is most like #823 but in comparing the two directly I prefer this one as being more realistic. I do think it has too much unused real estate in the midtones, so I would multiply it through a blurred layer mask based on the blue channel of the multiplied layer. 

828 Lee Varis has declared that he has become less of an LAB aficionado than in his youth and now prefers something like this version, as opposed to a fireball like #830, which he says he would have shot for after being introduced to LAB. I’m here to tell you that tastes do change over time but that certain concepts do not, including the idea of using full range every time. A hit of Auto Tone makes this a better version.

829 Another reasonable PPW effort making use of MMM, H-K, and channel blending. As this is a comparatively dark treatment, the greenery began to close up, so this person made the sensible choice of using Channel Mixer to lighten the green channel within the greenery.

830 Chosen for the par version. Like John Lund in #821, the person is familiar with the actual colors of the scene, but preferred what he called this “eye candy” version. He spent a lot of time experimenting with various method and blending them into his final; it included two separate hits of H-K to try to prevent the color becoming totally ridiculous. Some would find the final result over the top but I rather like it.

831 The par version, which I subsequently said unduly favored some violent versions like #830.

832 The “conservative par” version, not included in the original .zip file, based on five other versions that I identified as being good, but not spectacularly colorful in the sense that #830 is.







Re: Seeking Clarification of Procedure for Averaging Images

Rick Gordon
 

I haven't experimented yet, but I have a hunch that masks made from the L, and maybe even the a* and b* of a standard deviation stack rendering, maybe auto toned or maybe not, might offer some intriguing possibilities, maybe somewhat analogous to what can be done with auto toned difference masks.

Even the a* and b* channels of a standard deviation stack rendering show the contours of objects in ways that I bet, used as masks, could lead to some interesting results.

I've got to check that out.

Rick Gordon

--------------------
On July 9, 2020 at 4:17:35 PM [-0700], Thomas Hurd,md Via Groups.io wrote in an email entitled "Re: [colortheory] Seeking Clarification of Procedure for Averaging Images":
I expected a bit more from standard deviation.
___________________________________________
RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________
WWW: http://www.shelterpub.com


Re: Seeking Clarification of Procedure for Averaging Images

Thomas Hurd,MD
 

John, 

Thanks for showing that maneuver. I hadn’t tried it before, but it worked very well!

Obviously, the having fun with the the stack modes was just a little too much fun when I tried them. Maximum and minimum are good tricks to have. Imexpected a bit more from standard deviation.

Tom Hurd




On Jul 8, 2020, at 6:24 AM, John Gillespie <john@...> wrote:

You can make an average version by loading the images into a stack (File->Scripts->Load Files Into Stack). Select browse then click the images you want.
Once loaded (which may take a while if you choose a lot of images) select all the layers and then select Layers->Smart Objects->Convert To Smart Object (there is an option to do this on the file selection dialog but it doesn't always seem to work). Then select Layers->Smart Objects->Stack Mode->Mean.
You can also have fun with the other stack modes.

The alternative is to load all the images individually then adjust the opacity for each item, which needs to be done in proportion 1/N where N is the position from the bottom of the stack. So for 3 images it is 100% then 50% then 33%. With more than 5 images this is tedious and increasingly inaccurate as the percentage must be a whole number. So the first version is better if you are happy with a strictly equal contribution from each image. The second method is easier if you want to experiment with different weights.


Re: Monument Valley: comments on individual versions

Thomas Hurd,MD
 

Sorry for the partial post.

I’ve already outed myself as the owner of 827. At Dan’s suggestion in his comments, I did try the multiply though blue mask, and it worked just fine, better than a hue saturation adjustment layer.

 I’m not surprised Dan wasn’t sure what I did, because I just reread my description, and I could hardly understand it myself.

 I will try once more, maybe simpler.

 I blended three versions: 

Monday one was a quick mmmcb only, default values.

The second was a bit more detailed, with BIGGER HAMMER action and H-K before the MMMCB.

 Having a satisfactory version safely in my pocket, I went for broke Sunday afternoon and evening. 

I did a few unusual moves (for me) on the third version. 

  • After initial correction of the purplish cast, I tried to kill the red by blending the green channel into the red in a duplicate channel. For this layer I then used color mode. This was a little over the top so I had to add a curve layer to decrease the green a little bit in the mid tones.
  • This was correct in color, but I wanted to get the color variation compressed even more so I could use mmm and CB to my heart’s content when the time came. 
  • Also, I wanted to eventually increase the contrast to the max because there was no sharpening allowed and I really wanted to demonstrate the contrast between the reddish colors of the butte and the blue/gray.
  • So I used H-K and skin desaturation (it’s magenta and yellow so why not?)
  • Then (because some days I like to make pig of myself) I used S/H from the PPW Panel 6x in a row. (What the heck, I know I’m blending this with other images)

 

At last my colors dulled down, my contrast overdone with S/H, I tried something new (for me) the MMM fine-tuned action that Gerald Bakker developed as a refinement to the regular PPW. For the selection I drew a rectangle around the face of the butte. After the MMM fine-tuned (for those who haven’t tried it, you can find it as an action in “my actions” in the PPW Panel download), I used the color boost 2018, again from the same action panel.

  • Within the group that’s comprises the MMM-CB, I used the inverted yellow channel (gaussian blurred) to mask the color boost moves.
  • I made a composite layer.
  • I duplicated it several times.
  • The first layer up from the composite I multiplied on itself. This multiplies the L and overlays the a and b. Better color intensity and contrast.
  • The next layer I overlaid the a and b on themselves. I used a gradient mask to hide this from the lower half.
  • I inserted a curves layer, with the a and b curves flat across zero so there was no color. I decreased the opacity to about 40%, and made a gradient layer to hide this desaturation from the face of the butte.
  • I took the last duplicate and blended the inverted a channel into the L channel. This made the greens slightly lighter and the butte slightly darker to stand out from the background clouds.

 

I took this mountain of layers and made a composite layer.

I took the three versions. and blended them with color and luminosity % to taste.

 

The upshot is I really took more steps than I needed, but eventually got the luminosity changes I wanted, and blended the week’s worth of layers together.

 

Tom Hurd

 

 

 

 

Thomas Hurd, MD




On Jul 9, 2020, at 10:29 AM, Thomas Hurd <tomhurd@...> wrote:

I’ve already outed myself as the owner of 827.
I’m not surprised Dan wasn’t sure what I did, because I just reread my description, and I could hardly understand it.
I blended three versions: 
one was a quick mmmcb only.
The second was a bit more detailed.
Having a satisfactory version safely in my pocket, I went for broke Sunday afternoon and evening. 
I did a few unusual moves (for me) on the third version. 
After initial correction of the purplish cast, I tried to kill the red by blending the green channel into the red in a duplicate channel. For this layer I then used color mode. This was a little over the top so I had to add a curve layer to decrease the green a little bit in the mid tones.
This was correct in color, but I wanted to get the color variation slammed so I could use mmm and CB to my heart’s content when the time came. 
Also, I wanted to eventually increase the contrast to the max because there was no sharpening allowed and I really wanted to demonstrate the contrast between the reddish colors of the butte and the blue/gray.
So I used H-K and skin desaturation (it’s magenta and yellow so why not?)
Then (because some days I like to make pig of myself) I used S/H from the PPW Panel 6x in a row. (What the heck, I know I’m blending this with other images)
At last my colors dulled down, my contrast overdone, I tried something new (for me) the MMM fine-tuned action that Gerald Bakker developed as a refinement to the regular PPW. For the selection I drew a rectangle around the face of the butte. After the MMM fine-tuned (for those who haven’t tried it, you can find it as an action in “my actions” in the PPW Panel download), I used the color boost 2018, again from the same action panel.
Within the group that’s comprises the MMM-CB, I used primarily 

Tom Hurd

On Jul 8, 2020, at 9:46 AM, John Gillespie <john@...> wrote:

Pleased to be selected for the par version (812). Even if quite arbitrary, I'll take it!
One odd thing about the original is the vantage point. At first glance, as the foreground and the butte are so similar in colour and darkness, it looks like we are are seeing a horizontal foreground from which the butte emerges. But the tree on the left indicates that it is in fact a vertical bank and that the photographer is standing in a ditch or dried river bed.
Perhaps the ground dips again and the photographer is using this angle to obscure distractions such as other people, or a road etc. in the space between the bank and the butte. Or perhaps it is just to create some foreground interest.
Entry 827 makes the geography of the scene very clear (to my mind). You get a real sense that moving up the bank would take a bit of effort, and that there is a distance between the top of the bank and the main rock.
It may not accurately depict the colour of the foreground, but it does I think accurately show the shape of it more than any other, although 822 does a good job as well.
Blending with it definitely improves my version (or degrades 827 depending on your point of view).

 


Re: Seeking Clarification of Procedure for Averaging Images

Dan Margulis
 



On Jul 9, 2020, at 6:23 AM, Gerald Bakker <gc.bakker@...> wrote:

I don't want to draw many conclusions from just one example, but I noticed something that may be typical. Zooming in to 100%, the "mean" version looks smoother, whereas the "median" shows a bit more color noise and is slightly more splotchy. I think the reason is that averaging pixel values has a smoothing effect. Taking the median however implies that one pixel may come from version A and a nearby pixel from version B, causing a more rough appearance. The more differences between the base versions, the stronger this effect. 

Correct. Median is generally more useful when comparing gross objects, not individual pixels. If, for example, we wanted to know our consensus view on how colorful to make Monument Valley, well I suppose it would be possible to write a routine that evaluates how colorful each individual version is. In determining our consensus we would want to pick the median, because the mean would be too colorful (there are  more extremely colorful versions than extremely dull ones).

On a pixel level the noise occurs in transition areas. For example, everyone had different ideas about how to transition into the shadows of the greenery, some people basically just plugged them and others had a lengthy transition zone. In that transition zone, every change in pixel value from its neighbor affects the mean. It may affect the median but it may not. So the transition in the mean is smooth and in the median not.

Dan



Re: Seeking Clarification of Procedure for Averaging Images

Gerald Bakker
 

On Thu, Jul 9, 2020 at 12:01 AM, Rick Gordon wrote:
Based on the maxim that "the median is affected by outliers and skewed data than the mean, and is usually the preferred measure of central tendency when the distribution is not symmetrical" <https://www.abs.gov.au/websitedbs/a3121120.nsf/home/statistical+language+-+measures+of+central+tendency#:~:text=In%20a%20distribution%20with%20an,value%20is%20the%20middle%20value.&text=Advantage%20of%20the%20median%3A,the%20distribution%20is%20not%20symmetrical.>, that brings up an interesting note: as to whether doing this with Median as opposed to Mean would mitigate the damage of "one bad apple spoiling the apple cart."
That's an interesting thought for sure. I couldn't resist trying this out on one of the exercise images, the Panama photo, the only for which I had downloaded 5 versions. Well, the result surprised me, as the two procedures give quite different results. Different color ("mean" version somewhat more saturated) and different weight ("median" version a bit lighter).

I don't want to draw many conclusions from just one example, but I noticed something that may be typical. Zooming in to 100%, the "mean" version looks smoother, whereas the "median" shows a bit more color noise and is slightly more splotchy. I think the reason is that averaging pixel values has a smoothing effect. Taking the median however implies that one pixel may come from version A and a nearby pixel from version B, causing a more rough appearance. The more differences between the base versions, the stronger this effect.
--
Gerald Bakker
http://geraldbakker.nl

1081 - 1100 of 33174