The arsenal of weapons that we can deploy against recalcitrant images is powerful and impressive. When used prudently, that is. Assuming that because you have them you are guaranteed to create stunning imagery tempts you to try too hard to make it so. When this happens, as we’re about to see, the weapons, including PPW, are worse than useless. This post will talk about how to identify the images on which PPW is no better than alternate methods.

I have been posting PPW comparisons with a series of 150 images from the MIT 5k dataset. The opponents are five hired student retouchers, plus a sixth “par” version in which each of the five is worth 20%.

Because the images were taken at random, some will surely favor PPW over conventional methods. I wrote about those categories in the previous post, trying to suggest how you could identify images where you would expect good results. For example, pictures of deserts and canyons showed a great superiority for PPW.

The list of images where PPW has only a limited advantage or none at all starts with those whose main issue is incorrect color, such as an unpleasant cast. Other methods have tools just as powerful as PPW’s to correct these, although skill is required to use them properly. My own corrections on such images have a high win rate against the paid retouchers, but that’s experience, not system. If even one out of the five retouchers gets color as good as mine it suggests a limited superiority at best for PPW. More frequently, the par version, the average of the five corrections, turns out to be of good quality. Out of 150 comparisons between PPW and par, 17 were rated ties. Normally this was because the PPW version clearly had better contrast/detailing but the par version clearly had better color. We saw an example of this in an image of a young girl in this post.

4180-default: This image, although it might well be found in an advertisement, offers little chance for creative improvement. There is no reason to suppose that PPW would do better with it than any other method.

4180-default: This image, although it might well be found in an advertisement, offers little chance for creative improvement. There is no reason to suppose that PPW would do better with it than any other method.

The obvious category for which there is no advantage is the simple kind of picture with no technical issue and little room for creativity. Here’s one original, as received by the retouching team.

You’d never find so boring an image in a color correction class, but they do pop up from time to time in advertising work. Consequently its presence in the set, which is supposed to showcase typical images that a professional might encounter, is appropriate. I see no poimt in showing any of the corrected versions. I am losing little sleep over the fact that my entry did not win.

PPW has a big advantage when colors are subtle, as they are in images of canyons. It also does well when colors need to be intensified but not across the board. If neither factor applies we get to the hardest category to diagnose: when the scene contains many brightly colored objects and there is no reason to pay more attention to one than another. The four default images that follow invite the retouchers to create several strong colors. I invite you to predict which one(s) PPW will do better on. Getting the answer right is important, because I’m about to show, trying to force an advantage can backfire. I’ll show my versions next to the best of the rest, and explain how I would have answered the question before and after working on each.

2083-default and 2834-default: Both shots feature intense colors. Do you expect PPW to have an advantage in one, neither, or both?

2083-default and 2834-default: Both shots feature intense colors. Do you expect PPW to have an advantage in one, neither, or both?


4182-default and 4199-default: Two more scenes featuring bright colors. How will PPW do?

4182-default and 4199-default: Two more scenes featuring bright colors. How will PPW do?


We should agree at the outset that any reasonable workflow can produce bright colors on demand. If that’s all that’s needed there should be no advantage to PPW, unless one of the following things occur:
*The strongly colored objects are of similar hues that compete with one another.
*The strongly colored objects need to display more detail.
*The entire picture is so colorful that it is helpful to force more neutrality somewhere.
*Certain other colors need to be toned down to avoid competing with the strong colors.
*Certain colors need to be protected from getting more intense.

I wasn’t too clear on these rules when I started these exercises. but I did know enough to predict what would happen on the surfboard shot we’re about to look at.

4199-C: The best correction by any of the retouchers..

4199-C: The best correction by any of the retouchers.


4199-PPW: It's hard to say that this is any better than what the retoucher did.

4199-PPW: It’s hard to say that this is any better than what the retoucher did.


I do not show the average, the “par” result. because a couple of retouchers dragged it down with poor efforts. So my version wins over par, but not over the version of retoucher C. No apparent advantage to PPW.

This came as no shock to me. The surf equipment features lots of bright colors, but they are already well differentiated. The MMM action is therefore not particularly helpful. It’s moved some yellows toward orange, which is not a big deal one way or the other. The colors are basically blobs, not needing much detail, which wipes out another PPW advantage. And although the bright colors are striking, they are well distributed throughout. Most of the image’s real estate is dull and not particularly important, just the thing that makes bright colors pop.

The lack of these factors is why PPW is better on the following image of stuffed animals. This time, the challenger is the par version, which, as it often does, beat all five of the individual retouchers.

2083-par and 2083-PPW: The retouching team's averaged version, left, brought out the bright colors that are surely needed. But the objects lack the shape found in the PPW version at right.

2083-par and 2083-PPW: The retouching team’s averaged version, left, brought out the bright colors that are surely needed. But the objects lack the shape found in the PPW version at right.

Even though the original image was underexposed and muddy, the retouching team had no difficulty establishing the desired strong colors. But these objects are larger. They need more shape, unlike the surfboards. And the reds, yellows, and oranges compete with one another, a problem that did not exist in the previous image. Plus, a much higher area of this image is devoted to brilliant colors than in the other. When everything is colorful, then nothing is colorful.

PPW’s shape-creating tools, channel blending plus the H-K and MMM actions, attack these problems directly and create a more attractive result.

This extra shape is usually available in bright-color scenes. In the surfboard shot it was of no value. In the stuffed animals it was critical. A third category is where the extra shape makes a big difference but the question is whether it’s desirable, and if so, how much. The following image can be interpreted in several ways, so I offer two alternatives to PPW.

4182-par: This image can be interpreted in various ways. Unsurprisingly, the averaged version is correct, but conservative.

4182-par: This image can be interpreted in various ways. Unsurprisingly, the averaged version is correct, but conservative.


4182-B: This retoucher favored a warmer look.

4182-B: This retoucher favored a warmer look.


4182-PPW: The high-contrast look would be hard to duplicate without PPW. The question remains whether it is desirable.

4182-PPW: The high-contrast look would be hard to duplicate without PPW. The question remains whether it is desirable.

If these were submitted to a jury I expect that all three entrants would get votes. Mine, incidentally, goes to retoucher B. My idea was to make the background very dark, hoping that this would make it seem that a spotlight was shining on the girl. It did that, and it would have been hard to duplicate this version outside of PPW.

Some might like this overdramatic treatment. Unlike the other two, which are bland enough, I can see someone actively disliking mine. It’s another reminder that when we get a good idea it is easy to fall in love with it and take it too far. Still, I expect that those who favor one of the other two versions would concede that each would be helped if it moved slightly in the direction of mine. Without PPW, that option was not open.

The final example reiterates what I said at the outset about the dangers of trying to force something to happen.

2834-par: Creating brightly colored clothing is not a problem.

2834-par: Creating brightly colored clothing is not a problem.


2834-PPW: A misguided attempt to tone down all but the brightest colors, a move that worked well in the stuffed animals image.

2834-PPW: A misguided attempt to tone down all but the brightest colors, a move that worked well in the stuffed animals image.


My wretched performance on this exercise came because I decided it was analogous to the stuffed animals image, which was in danger of being overwhelmed by bright, featureless colors. The solution there was to darken and desaturate areas that were colorful yet not the most colorful things. The move added shape and made the strongly colored objects stand out more by comparison.

Trying the same approach in this clothing image was a poor idea. Although the bright colors dominate in terms of jumping to the eye, a large percentage of this image is not colorful at all. There was no point in toning those areas down. All that doing so accomplished was to make it tough for me to get to the attractive bright colors of the par version, colors that needed no help in being distinguished from one another.

We close on this painful note, hoping that it reminds us that when there is no reason to believe that the fancy way has an advantage, stick with the straightforward way.

{ 0 comments }

Almost any method of correcting images works some of the time. For those interested in PPW, or in taking a four-day class on how it works, the question has to be how much of the time. The MIT study we’ve been looking at offers a unique opportunity to answer that question. It shows the real-world advantages of the PPW and its various components, most of which I knew already, but there were a few surprises. I will discuss the two categories in this and the next post.

My book on the uses of LAB in Photoshop is subtitled The Canyon Conundrum. The first chapter shows corrections of nothing but canyons. Back in the first edition, I listed all the Photoshop books of the time that briefly mentioned what LAB was. Each one illustrated the point with a shot of a canyon.

As I pointed out then, one might conclude that working in LAB has major advantages for canyon images. Since the Picture Postcard Workflow uses LAB as its standard way of bringing out color, it follows that PPW also is a superior way of working with canyon images. For example, from the MIT 5k dataset, here is a “par” correction, compared to a PPW version prepared by me from the same original.

0018-par: The averaged result of five retouchers,

0018-par: The averaged result of five retouchers.


0018-PPW: A PPW correction based on the same original.

0018-PPW: A PPW correction based on the same original.

It is a trap to believe that since PPW is greatly superior on this image it ought to be similarly superior on whatever the next one is. That next one might not be a canyon. Human nature ensures that when we believe we have invented a better mousetrap, we try to prove that it is decisively to existing mousetrap technology in every conceivable circumstance. In our field, some misguided people (and I am one of them, far too often) try too hard to get something spectacular when the competition’s method works just as well.

One question is, how often does the competition have methods just as good? That’s hard to know from what I’ve written in the past. I save the images in which PPW does well for demonstrations and articles, but I basically ignore the images in which it does not have an advantage. That’s why the MIT study is so valuable: if we pick enough of the 5,000 MIT images at random we’ll get our fair share of both varieties. Originally I thought 50 images would be sufficient, but even when I raised the number to 100 it seemed like it might be somewhat biased toward images that favored PPW. So, I went up to 150 randomly chosen images, which fall into the following categories.

44 mostly people (29.3%) in the following subcategories:
images dominated by faces 7 (4.6%)
moderately large faces 13 (8.7%)
smaller faces/full figure shots 24 (16.0%)

30 scenic shots (20.0%) in the following subcategories:
greenery predominates 11 (7.3%)
desert or canyon 10 (6.7%)
other scenic 9 (6.0%)

14 studies of animals or birds (9.3%)
8 night or twilight shots of cities (5.3%)
15 architectural shots (10.0%) in the following subcategories:
interiors 7 (4.7%)
exteriors 8 (5.3%)

8 studies of flowers (5.3%) in the following subcategories:
basically all one color 4 (2.7%)
multiple colors 4 (2.7%)

5 studies of a strongly colored object other than flowers (3.3%)
5 food (3.3%)
3 sports (2.0%)
20 miscellaneous (13.3%)

As described here, each MIT image was corrected in Lightroom by five compensated student retouchers. I created a sixth “par” version, in which each of the five results was weighted 20%. Because the par version was often better than any of its five parents, I was most interested in comparing the par to the PPW version. But I also compared both the par and the PPW versions to each of the five parents, limited only to the question of whether any of the five were competitive with, or even better than them.

In evaluating the results (ground rules described here) it must be conceded that the MIT retouching team was made up of persons who, though reasonably skilled, don’t have nearly as much experience in the field as I do. So in trying to determine how often PPW makes a major difference I look at all of the cases where the PPW version was decisively better than the par version, but excluding ones where at least one of the five retouchers created something competitive. Then, I apply some kind of mental fudge factor downward on the theory that these retouchers are more likely to do something foolish than I am.

I’ll show the full results at the end of this post, but the bottom line is that PPW appears to add significant value to slightly less than half of these images. Certain categories are quite predictable. Ten shots of deserts and canyons? I’d expect a clean sweep, and that’s what happened. Of 50 individual and 10 par versions, only one retoucher managed to get something competitive to PPW.

One must expect similar results in scenic shots that feature greenery, due to the MMM action in PPW, and sometimes H-K as well. For example,

4197-par: The averaged result of five retouchers.

4197-par: The averaged result of five retouchers.


4197-PPW: A PPW correction based on the same original.

4197-PPW: A PPW correction based on the same original.

Other types of scenics also favor PPW. In this post I showed an image of Waikiki Beach where the extra action in the water was a big plus compared to the work of the Lightroom retouchers. So far, this is a good explanation why more landscape photographers seem to prefer LAB than anyone else.

As expected, PPW also did well in two other categories. When an image is dominated by one object of a single color, or in any of the flower studies that featured only one color, the PPW blending capabilities, sometimes assisted by H-K, created a decided advantage, as in

4537-par: The averaged result of five retouchers.

4537-par: The averaged result of five retouchers.

4537-PPW: A PPW correction based on the same original.

4537-PPW: A PPW correction based on the same original.

The final category in which one would expect a near clean sweep for PPW is in shots of faces. The MMM action creates attractive hue variation within the face, something missing in other applications.

3498-D and 3498-PPW: Left, the best of the five retoucher versions, which was better than the averaged entry. Right, the PPW version.

3498-D and 3498-PPW: Left, the best of the five retoucher versions, which was better than the par entry. Right, the PPW version.

This category offered two surprises, at least to me. First, I thought that the advantage would hold even when the faces were fairly small. It did not. Medium-sized faces generally were better with PPW, but it wasn’t as decisive as when the faces were larger. And even with larger faces, an exception showed up: African-American flesh is not as flattered by hue variation as lighter skintones are.

The categories listed above comprise about a third of all the images. PPW often has advantages in the others but they are image-specific as opposed to being common to a category. For example, PPW has a high win rate in architectural images because its false profile/multiply routines can correct for poor lighting conditions. Also, PPW has excellent handling of shadow and highlight detail, which are often big factors in images of birds and animals. But if these factors don’t exist then if you use PPW virtue will have to be its own reward.

Sometimes these factors pop up in an unexpected category, such as night or twilight shots of cities. In principle there isn’t a PPW advantage because the main decisions (how dark to present the scene, and what color for the sky) are artistic choices for which technique is irrelevant, and because the bright lights in the buildings are easy to handle in any method.

An image of the famous fountains of the Bellagio in Las Vegas, however, is the exception.

4196-par: The averaged result of five retouchers.

4196-par: The averaged result of five retouchers.


4196-PPW: A PPW correction based on the same original.

4196-PPW: A PPW correction based on the same original.


True, there is plenty of room for debate over the color of the lake and sky, and how strongly and what color to make the lighting of the building. But the signature of this hotel is its magnificent array of fountains and this shot lives or dies by their detailing, which PPW’s Bigger Hammer handles well.

Here are the overall results:
Comparisons against the five parent versions:
Par version is significantly better than a given parent: 558 times out of 750. (74.4%)
Par version is significantly better than ALL FIVE parents: 36 times out of 150. (24.0%)
PPW is significantly better than a given parent: 706 times out of 750 (94.1%). On only two images were there three parents competitive to PPW; in three others two parents were competitive.

PPW vs. Par (150 comparisons)
126 total wins (84.0%) with the following subdivisions:
Decisive wins over par; significantly better than all parent versions: 72 (48.0%)
Decisive wins over par but at least one parent was competitive 16 (10.7%)
Significant wins over par and all parent versions 31 (20.7%)
Significant wins over par but at least one parent was competitive 7 (4.7%)
17 ties (11.3%)
7 total losses (4.6%) with the following subdivision:
Decisive losses 2 (0.7%)

{ 1 comment }

For those wishing to take color skills to the ultimate level, here are the two dates for Applied Color Theory classes in 2018.
• ATLANTA, Wednesday, May 16, through Saturday, May 19.
• SAN DIEGO, Wednesday, August 22, through Saturday, August 25.

These classes—four long days, limited to eight persons—have changed the lives of many a photographer, many a retoucher, and many a Photoshop authority. This will be the 24th year I’ve been teaching them, in a dozen different countries, in four different languages. And never the same way twice. Every group has its own character, faces its own challenge, exults in its own successes. I never got tired of it, because I never found it repetitious.

Of course, the class has changed drastically over time, adding new techniques and interesting imaging problems. In 2011 we switched from three to four days, in view of the increased complexity in our field. But the format is the same. I teach what I can, then the group works independently on a set of images, and then the results are compared. This routine is repeated seven more times over the three days. By the time the class is over, we’ll have compared our results on 28 different images. The first 12 are the ones that every class works on. The last 16 are customized to and chosen by the class. Often they include images that class members themselves have provided as representative of their own work. This year, we’ll be adding some of the images from the MIT set that I’ve been writing about in this blog.

All of us who are serious about the topic have developed our own methods which seem to work well for us. Unfortunately, we normally have nothing to compare the results to. Many methods, some quite crude, still can make original photographs look a lot better. The question always is, how much better could it have been?

Nothing can answer that question better than seeing what others can do with them. In principle, you ought to be able to get what you think is the best result, because nobody else can read your mind to know what you’d like to do with the picture. In practice, especially during the first two days, there is a lot of wailing and gnashing of teeth from students whose work did not measure up to their own expectations, in comparison to that of others.

Four days, 40 hours. It remains the fastest way that a professional—or someone who wishes to produce professional quality work—can upgrade color correction skills. If you’d like more information, or to consider signing up, here’s the link.

{ 0 comments }

White Point, Dark Point, Auto Tone: The Simplest Move of All (The MIT 5k Dataset 5)

by Dan Margulis December 10, 2017

“And yet,” I wrote in my first book 25 years ago, “most color correction could be handled by monkeys…a numerical, curve-based approach calling for little artistic judgment…all the advanced techniques are inevitably based on these surpassingly simple ones. The by-the-numbers rules can be stated in a single sentence: Use the full range of available tones […]

Read the full article →

The MIT 5k Dataset 4: More on Averaging

by Dan Margulis November 28, 2017

The previous entry described giving each of five independently corrected versions 20% weight to create a new, “par” version. This can be called a “stupid” blend, in that no notice is taken of the merits of any of the five. Nevertheless, it appears that this average is better than all five of its parents in […]

Read the full article →

The MIT 5k Dataset 3: Effective Averaging Close-Up

by Dan Margulis November 18, 2017

Those interested in quality have always been willing to spend time to get what they considered the best possible results. For some years now I have been suggesting that this is not the best approach in our field. Instead, I have been preaching that it is a better use of time to do the initial […]

Read the full article →

The MIT 5k Dataset 2: The Ground Rules

by Dan Margulis November 13, 2017

The following details the procedures used in evaluating the images in this study. It is posted separately so that I do not have to repeat it every time I discuss results in the future. I went through the set of 5,000 original images and deleted those I thought were of limited interest. I used the […]

Read the full article →

The MIT 5k Dataset 1: Introduction

by Dan Margulis November 11, 2017

This is the introduction to a series of posts I will make based on my work with files that are part of a remarkable archive. Researchers at MIT and Adobe have recently made available the data from a massive project they have undertaken to study what people look for when they correct color. The researchers’ […]

Read the full article →

Applied Color Theory classes in 2017

by Dan Margulis January 4, 2017

For those wishing to take color skills to the ultimate level, here are the two dates for Applied Color Theory classes in 2017. • ATLANTA, Wednesday, March 22 through Saturday, March 25, 2017. • SAN DIEGO, Wednesday, August 9, through Saturday, August 12, 2017. These classes—four long days, limited to eight persons—have changed the lives […]

Read the full article →

The Presentation of Data: When Red and Blue Are Opponent Colors

by Dan Margulis June 3, 2016

The U.S. presidential campaign offers an interesting insight on opponent colors, and on how best to present data. The conventional way of doing it leaves much to be desired. Residents of other countries have difficulty understanding the American system, where in effect the election is always decided by voters in a small minority of states. […]

Read the full article →