Monday, February 11, 2019

International Day of Women and Girls in Science

Today is, of course, the International Day of Women and Girls in Science. I should clarify, As I write this post, it is the International Day of Women and Girls in Science. As you read this post, it may no longer be the International Day of Women and Girls in Science. This is sad, because the International Day of Women and Girls in Science should go on forever, to remind us that Science (with a capital S) is not just about guys, but it's about guys and gals. I personally would like for Science to not be just a men's club.

If you are a blog post aficionado, and you are spending your International Day of Women and Girls in Science reading blog posts about the International Day of Women and Girls in Science, I'm gonna guess that you are getting tired of reading about Marie Curie. I mean every blogger and his/her left-handed third cousin twice removed from Akron is writing today about Marie Curie. I certainly don't want to belittle her contributions to Science. I mean, there's that whole Nobel Prize thing. I guess that's kind of a big deal.

But I do want to highlight someone else, someone who has not gotten the recognition that she deserves.

My first thought was Ada Lovelace, who was the very first computer programmer. She programmed a computer developed by Charles Babbage, the Analytical Engine. She programmed the computer before it even existed. How cool is that? I'm not sure how she debugged her code.

A model of Babbage's Analytical Engine which has been on backorder since 1837 

Naaaahhhh... she's been overdone. They even named a programming language after her. Not "Lovelace". That has to do with a film that no one will admit to having watched. The language was Ada.

Then there's the early programmers of the ENIAC -- a computer which actually existed. Likely becasue of the war, these Rosie the Riveters shown below were the first to program the ENIAC. The programming on this beast wasn't done by typing in stuff like "If x = b, then ...". It wasn't even done by flipping switches for 0s and 1s. It was done with short cables that connected various arithmetic units together.

Kay McNulty, Betty Jennings, Betty Snyder, Marlyn Wescoff, Fran Bilas and Ruth Lichterman were the original programmers of the ENIAC

My next thought was Florence Nightingale. She deserves some credit, since she doesn't get the credit that she deserves. I think most people would remember her as a caring nurse at a field hospital during the Crimean War (in Turkey, circa 1856). This is true. She was a nurse to the soldiers. But to call her "just a nurse" significantly underplays her contribution to Science.

As a nurse, she saw many soldiers die, and was understandably concerned. But rather than just being compassionate, or complaining about the constant deaths of the wounded, she did something about it. She took data. She drew some cools info-graphics. I have no idea what the chart below shows, but someone told me that it is a visualization of the mortality data by month.

Early infographic on field hospital mortality

In the end, Nightingale determined that unsanitary conditions were a major deciding factor into whether a wounded soldier survived. I see this as a three-fold win. First, of course, there was the immediate short-term benefit of saving lives. And that's what war is all about, isn't it?

The second benefit is one that is clear even today when someone goes into surgery. The area of the incision on the patient is carefully cleaned. The surgeon washes his or her hands very thoroughly, and is careful not to get the hands dirty after that. Even in every hospital room, we see a hand sanitizer dispenser on the wall.

The third benefit is more nebulous. From the fact that Nightingale is remembered as a nurse, I conclude that this point may have been lost on the general public. The nebulous and under-appreciated learning from this story is that data can drive decision making. She took data, she analyzed data, came to conclusions based on that data, and those conclusions led to significant improvement to the process.

To quote William Deming, an early proponent of statistical process control, "In God we trust. All others bring data."

Then I come to the woman who I decided would be the topic of this blog post, Dorothy Nickerson. She started her work at the Munsell Corporation and went on to work the bulk of her career at the US Department of Agriculture. She was instrumental in the development of measurement of color, process control in the manufacturing of color, and development of international standards for color.

This looks like a good paper!

I wish I had started this post a few days ago, cuz I would have done a bunch of research on her and presented a really cool and compelling story. My apologies. I promise to get to that blog post some time in the future.

Wednesday, January 30, 2019

The Color Name Conundrum

This article was flagrantly stolen from my keynote presentation at ISCC/AIC Munsell Conference, July of 2018, and from the ISCC Newsletter of January, 2019.

It’s a common argument that my wife and I have. We are at a store or movie or coffee place, and I will comment on another woman’s blouse. “Hey, Honey. Look at the woman in the turquoise top. Isn’t she cute? … She smiled at me… And she handed me a card with her number on it.” Madelaine will invariably respond with “That’s not turquoise!” She may say that it’s teal, or aqua, or beryl, but she will never agree on the color name that I chose. I can blather on all I want about how I am a world-famous color scientist who was asked to give a keynote for the Munsell Conference. It won’t matter. What do I know about color?

The lady in the allegedly turquoise top

This time, I decided that I would win the argument. I started with Merriam-Webster’s dictionary since it is an authoritative reference that would show I was using the color name correctly. This dictionary defines turquoise as “a bluish-green color”, and follows up with the full and much more explanatory definition “a light greenish blue”.

I exercised due diligence and spoke directly with the person who wrote the full definition, Kory Stamper, to help resolve the argument with my wife. She politely (and wisely) declined to get involved. But I could tell that she was agreeing with me.

[As an aside, the exciting thing about attending ISCC/AIC Munsell conferences is that eminent chromo-lexicographers like Kory might be in the audience when they are called out in a keynote address.] defines turquoise similarly: “a greenish blue or bluish green color”. The Oxford English Dictionary provides a similar definition but leans more to the greenish side: “a greenish-blue color”. So, it seems we have a consensus between the dictionaries. But more importantly, we have a consensus in which I win the argument!
The image below shows blue, greenish-blue, bluish-green, and green. The blouse is definitely close to bluish-green, so turquoise is indeed an appropriate descriptor of the blouse color. Did I mention that I claim victory?

The happy shades between blue and green

But I decided to check one last dictionary, Webster’s Third New International. The definition in this dictionary is at once beautiful and tedious.

1) a variable color averaging a light greenish blue that is deeper and slightly greener than average turquoise blue, and greener and deeper than average aqua or average robin’s-egg blue (sense 1)

My stalwart research assistant suggests that the definition might be a bit too complicated

You can see that our puppy, Mozart, was puzzled when he read it, so I diagrammed the definition out for him (see next image). He thanked me when he saw the diagram, and went off to bark a friendly greeting to a squirrel that was outside. By the way, Mozart is not named for Hank “the Tank” Mozart. You will recall that Tank played defensive hatchback for the Green Bay Bruins. His claim to fame is that he scored the winning basket over Jack Nicklaus in the 1968 War of the Roses Tournament. Madelaine and I named the dog after the less-well-known Wolfgang “Wolfie” Mozart.

An Applied Math Guy reads the dictionary

In most dictionary definitions, the lexicographer works to define complex words in terms of more basic words. The Webster’s Third definition of turquoise is unique in that it defines the color relative to other colors which are just as non-basic as turquoise. To really make sense of this tortuous definition of turquoise, I realized that I had to generate similar diagrams for aqua and robin’s egg blue and turquoise blue and greenish-blue, and then for each of the other colors that were called out in those definitions. It only took me three days to generate the following table that delineates the territory of the ten tones in the turquoise tautology. It is clear from this that color names are very precisely defined.

A handy reference for color names in the blue-green family

But I still wasn’t happy. The intertwined definitions haunted me. Where Kory is the Steinbeck of chromo-lexicography, whoever wrote the lovely and sadistic color definitions from Webster’s Third was the Faulkner. I simply had to find out who this anonymous author was.

Luckily, it didn’t take long. The list in the front of the dictionary of contributing experts provided me with the answer. It had to be Isaac Godlove.

[As an aside, the exciting thing about attending ISCC/AIC Munsell conferences is that the audience will recognize the names of prominent researchers in color when their names are mentioned in a keynote address. Let me tell you, the cheers were deafening! Everyone recognized that Godlove was the third author of the seminal paper “Neutral Value Scales. I. Munsell Neutral Value Scale” from the Journal of the Optical Society of America in 1933.]

Of course, some of the people cheering also recognized that Godlove was the director of the Munsell Research Laboratory from 1926 to 1930. What an enormous coincidence that he should get mentioned in the keynote at the Munsell Conference! A few chromo-historians in the crowd actually knew that Isaac Godlove was the chair of the ISCC Committee on Measurement and Specification in 1933. (Note again the coincidence that the ISCC was one of the organizers of the Munsell Conference!)

While Godlove was chair, a group of pharmacists approached Godlove about the need for a definitive guide to color names. This eventually led to the National Bureau of Standards runaway best seller “Color – Universal Language and Dictionary of Color Names”, which became a Broadway play of much acclaim. This absolutely delightful standard carved the Munsell Color Space into 267 regions (called Centroid Colors) and gave each region an intuitive designator like “bG 159”, along with a euphonious name like brilliant bluish green.

A hue slice from the NBS standard on color names

As if that wasn’t enough to earn a prominent spot in my bookcase, the authors dug through all the available color naming guides (like Maerz and Paul, Plochere, and Ridgway) to determine the Munsell coordinates for each of the color words that were defined. As a result, the NBS standard further provides two lists: 1) a list that goes from common color name to the appropriate Centroid Colors in Munsell space, and 2) a list that provides all the color names that have been associated with each of the 267 Centroid Colors.

I was ecstatic. I quickly saw that this book provided a solution to the recurring argument that I had with my wife. The solution is astoundingly simple. Whenever I am within earshot of Madelaine, I just have to go through four simple steps before I utter any color names.

Step 1: Measure the color in question. For example, I called up the woman in the turquoise top, explained the situation, and met her at Starbucks with my spectrophotometer so I could measure her shirt. She understood my predicament perfectly, and agreed to share a Starbucks with me. Her shirt measured CIELAB of 86, -47, -4. Her name is Teal, by the way.

Yes, it’s a bit of a bother for me to carry a colorimeter with me at all times, but what color scientist worth his or her salt doesn’t carry one for the occasional color measurement emergency?

Step 2: Convert from CIELAB coordinates to Munsell designation. One could make use of the Munsell Renotation Data. The official version is conveniently available on the RIT website to do the approximate conversion, but several people have written software that does this. Harold Van Aken (of Wallkill Color) provided a piece of software as a freebie in honor of the Munsell Color Conference. (Yet another astounding coincidence.) Paul Centore has graciously provided an open source conversion, and Danny Pasquale sells an inexpensive tool called PatchTool that provides this function among others. The CIELAB coordinates of Teal’s allegedly turquoise shirt were thus converted to 5BG 8.5/9 in Munsell notation.

[As an aside, the exciting thing about attending ISCC/AIC conferences is that two of the three people who wrote software for this conversion (Paul and Danny) were actually in the audience for the keynote.]

Step 3: Convert from Munsell designation to Centroid colors. It goes without saying that it is pretty quick and easy to leaf through the diagrams (like the one below) in the NBS standard to find the Centroid corresponding to any Munsell designation. In this case, the Centroid Color is 159. Yes, it’s a bit of a bother to carry the NBS standard with me, but it’s a small price to pay for me to prove that I am right in an argument with my wife.

Step 4: Look up the color names listed under the Color Centroid. In the case of Centroid 159, the list is rather short. It includes Beryl Green, Bewitch, Blue Green, Bluish Green, Bright Aqua, Bright Aqua Green, Bright Emerald Green, Bright Green, Bright Jade Green, Bright Turquoise, Bright Turquoise Green, Chill, Crest, Du Barry Blue, Festival, Green, Ice Boat, Light Emerald Green, Lilting Green, Naid, Persian Green, Picturesque, Pool Green, Promised Land, Salome Blue, Song of Norway, Sprite, Sulfate Green, Turquoise Green, Venetus, Venice Green, … and of course, Turquoise. I win!

The fact that this particular color has 32 valid names shows that our assignment of color names to physical colors is not nearly as precise as Godlove and Webster’s Third would have us believe. We need a system like Munsell or CIELAB (or NCS or RAL or Pantone) in order to accurately communicate colors. That’s an important thing to realize, but the more important takeaway from the research presented here is that I won the argument!

May you enjoy arguing with your significant other as much as I do.

If you enjoyed this article, you might consider joining the Inter-Society Color Council! Individual membership is only $50 per year, for which you will receive the ISCC newsletter, as well as reduced rates for any ISCC sponsored conferences.

Such as... the joint TAGA/ISCC conference in balmy Minneapolis in March.

Wednesday, January 23, 2019

Which way is north in Munsell color space?

I wrote a blog post for Inkjet Insight about the Munsell color space. I don't want to spoil it for anyone, but the post is mostly just a gateway post to one about the CIELAB color space. For the Inkjet Insight post, I had my crayons pose for the aesthetically pleasing picture below. I do expect an Emmy for the picture, but I will try to appear surprised when I get called to the stage.

Happy crayons get together for a crayon picnic

Boy! Did that picture stir up a hornet's nest when I posted a link on LinkedIn! Two of my color scientist friends took umbrage. You may be wondering about my choice of the word "friends". Perhaps I use the term loosely, but Danny and Dave are the closest thing I have to friends, I mean, aside from Truffle and Mozart. And I feed Truffle and Mozart twice a day.

Here's Danny's malicious comment: "It seems that you have ordered the crayons as CIELAB would but not as Munsell does."

Dave's equally viscous comment: "Oops. Danny is right of course.  Unless, ... this is a view from below! To be more CIELABish I actually reverse the hue direction in my hanging Munsell Tree."

The gauntlet has been thrown down!!

I gotta ask you gentlemen, Danny and Dave, which Munsell color system you are referring to?

First, there is nothing inherently in the Munsell notation (7.5PB 4/6) that tells us which color points east (0 degrees) and whether orange is clockwise or counterclockwise of red. The Munsell notation for each color includes one of ten designators (R, YR, Y, GY, G, BG, B, PB, P, or RP) to specify a hue family. Within each hue family, there are ten steps which (oddly enough) are numbered from 1 to 10. Each number is one-step change in hue. Thus, 7.5PB is a unique specification for a hue, without any implied orientation.

Take that, Danny and Dave!

Second, there is a disagreement between Albert Munsell and Albert Munsell about the direction of red. As shown in the image below, his 1915 atlas has red pointing at around 45 degrees clockwise of east.

Page ripped from the Munsell Color Atlas of 1915

But in Munsell's New York Times 1919 bestseller "A Color Notation System" the master shows red pointed due north.

More vandalism, but to Musell's A Color Notation System

Both of these Munsell illustrations show orange as being counterclockwise from red. Pretty much the same as the way my Crayolas arranged themselves, and also, the way that CIELAB is arranged. Since there is no "correct" direction for red to point, I feel justified in pointing red to the east.

How do you like the color of them apples, my Dynamic D-named Duo!?!??  Well, I'm not done yet!

Third, the ASTM disagrees with both of these Munsell orientations. In 1968, the ATSM provided us with a "standard method of specifying a color by the Munsell system" (ASTM D 1535). Note that in the ASTM system, red points to the north. This agrees with the second illustration, but hang on a sec while I expound on some ASTM D 1535 trivia that is likely to come up the next time Danny, Dave and I get together for sushi.

ASTM D 1535 dictates this orientation

ASTM D 1535 also assigned a number for each discrete step of Munsell hue angle, from 1 to 100. Interesting point -- their notion of hue "angle" is in centicircs. I just made that word up. One centicirc is 3.6 degrees. It's about time we went metric and got past this silly Babylonian notion that we should measure arcs by comparing against the size of arc that the Earth makes around the Sun in a day. Approximately a day.

The ASTM adopted the obvious convention that 0 Munsellian centicircs would be at 18 degrees counterclockwise from north. I mean... of course. Well... I need to clarify. No one ever really told me whether "true" red was 0R or 5R. I guess I assumed it was 0R.

Also D 1535 is not explicit, but I think that 0 centicrics is not allowed. That has to be called 100. Kinda like the zero-phobia that says that midnight is 12:00 instead of 0:00. And that the first day of a month is 1, rather than 0.

More importantly, note that contrary to "normal" analytical algebra, Munsellian centicircs increment in the clockwise direction! I'm going to report them to their calculus professor!

But hang on. Here's the big thing. Even more importantly, and contrary to Munsell's two books, in the D 1535 system, orange is clockwise from red. O.M.G.!!

Dave boastfully mentioned his "hanging Munsell Tree". Not to be outdone, I provide a picture of my own hanging Munsell Tree. In my case, the world famous Munsell Color Model is joined by the world famous Munsell Color Model Model, Madelaine. We can see that my tree and Dave's tree both adhere to the "orange is counterclockwise from red" convention.

Wild times at the John the Math Guy household

Since there seems to be some latitude in the orientation of the color wheel in Munsell space, I claim that I am well within my rights to orient Munsell space in such a way as to serve as a stepping stone to CIELAB.

For those of you who are wondering about the significance of all this detailed historical research, let me be clear. It's all about me proving that I am right. Nothing else really matters.

NOTE: I would like to thank Robin Myers for pointing out an egregious error in my initial post. I had stated that the Munsell Color Atlas was published in 1913. Robin sent a photo of a page from his very own copy of the Munsell Atlas that clearly shows the date as 1915. (I am so jealous that he has this copy!) I would make up some excuse for why I got this wrong, but it would either be a total "my dog ate my homework" excuse, or it would make me look bad. So, I will just apologize for any pain and suffering which may have been caused by my ineptitude. I am eternally grateful to Robin for finding this embarrassing error, and would like to publicly offer to buy him a cup of coffee or a glass of the most inexpensive beer that can be found, provided he reciprocates by buying me a drink of similar value.

While John the Math Guy, LLC strives to maintain the highest level of scholarly eptitude in all its blog posts, there will inevitably be lapses into complete failure of logic, due diligence, and clarity of exposition. Any liability for anyone actually taking any of these posts seriously shall be limited to any considerations received directly from the party who has his undies in a bundle about this stuff.

Tuesday, August 21, 2018

State of the art in Extended Gamut printing

I had a series of blog posts on expanded gamut (parts one, two, and three) which were very popular. When I say "very popular," I of course mean that I have indirect evidence that there may have been one person -- possibly in Spokane, WA -- who stayed on the web page for longer than one minute. While it is likely that he or she was not actually looking at the screen at the time, we cannot dismiss the possibility that someone actually read part of one of the previous blog posts.

Obviously, I need to follow up, to provide some practical advice on expanded gamut. Since I am only capable of impractical advice, I have called on a friend who has convinced me that he knows stuff about this stuff. I turn this over to today's guest blogger, Mike Strickler.


John has treated us to an entertaining history of attempts to achieve more colorful results by overcoming the limitations of 4-color printing. But what of the situation today? What does “Extended Gamut” mean in the present context; what do these solutions look like? EG systems now range from simple arrangements consisting of nothing more than Photoshop and a multichannel output profile to entire integrated workflows with proofing and elaborate options for spot color handling. But for all the recent attention paid to the subject, there is still a lack of industry consensus on just what a constitutes a proper EG separation, and any effort to make sense of the subject typically faces a mixture of conflicting proprietary claims, incomplete studies, and persistent misconceptions. It is still the Wild West. Perhaps by considering how these systems came to exist we might better understand what they actually do and how well they fulfill their purpose.

Extended-gamut today: a dual heritage

We can simplify the origins of all EG systems to two distinct lineages, each corresponding to a different need: Systems that convert images and those that convert spot colors. Combining and reconciling these two functions is the key challenge faced by designers of EG systems today.

Spot color conversion systems: a tyranny of rules

As we read earlier, several decades ago some clever individuals saw that one might reproduce a wide range of spot colors as equivalent process builds using CMYK with the addition of secondary inks such as a red, orange, green, blue, or violet. This was in the days before digital color management, so lookup tables were laboriously built from trial press runs and served as guides for converting colors, object by object. These systems mostly used AM screening, so immediately the question of screen angles had to be addressed. The usual scheme of 0, 15, 45, and 75 degrees was seen as imperfect because it forced a choice of either creating additional intermediate screen angles, increasing the risk of moirĂ©, or sharing screening angles among complimentary colors such as cyan and red/orange and magenta and green, an unpopular choice as it raised the risk of color shifts caused by misregistration on press—a reasonable concern when printing vector objects, with their well-defined boundaries. A near-consensus converged around two simple rules for multicolor separations:

1. No color shall be converted to more than 4 process colors. Three is even better.

2. Complimentary overprints, such as cyan-red/orange and magenta-green, must be prevented. Who needs them, anyway?

These rules continue to be applied in a majority of EG systems, even those updated with automatic color-managed conversions. Spot color tints and overprints are handled by arithmetic interpolation or other simple means. The intricacies, and limitations, of these schemes will be discussed in the next blog. For the remainder of this one, we’ll focus on image conversions.

Images conversion systems: the rise of multichannel ICC profiles

While those brave pioneers were building their lookup tables, deft prepress workers were enhancing CMYK images with “touch” or “bump” plates of stronger colors. This practice can be seen as the true progenitor of multichannel extended-gamut systems.

With rare exceptions, modern image conversion methods rely on ICC color management. For generating multicolor (CMYK+N) press separations a basic system includes two profiles, one for the source color apace (often RGB) and one for the multicolor destination space, a CMM (Color Matching Method), AKA a “color engine,” and an application such as a RIP, color server, or other color-managed program such as Photoshop to interpret the source pixels and build the new image.

The profiles contain lookup tables that translate color appearance values (XYZ or Lab, AKA the PCS, or Profile Connection Space) to device values and vice versa. The CMM draws smooth curves through the LUT points and interpolates all intermediate values. PCS to device (or B2A) tables may contain a good deal of “secret sauce” for enhancing printability, saving ink, increasing the amount of black in neutrals, etc.; separate tables are built for different rendering intents or gamut mapping strategies. On top of this, device-link profiles may be employed to bypass the PCS conversion and apply even more rules, e.g., to exempt certain colors from the conversion. The key points to know about multicolor ICC systems are:

They are automatic, fast, and precise.

They are able to juggle multiple objectives, the two most important being accuracy and smoothness of output color, even up to 7 or more output channels.

Any combination of output channels may be used to fulfill objectives.

 Results are conditioned by the quality and completeness of underlying measurement data.

Two systems, two outcomes

Image conversion is still an important function of EG, and systems do perform differently depending on their underlying logic. To illustrate this we can look at the results from converting an RGB test image with two very different solutions, one developed primarily for simulation of spot colors—you might say an heir to those clever lookup tables—and the other a typical representative of ICC color-management, the current default method for converting images. The test image contains a smorgasbord of truly devilish color conversion challenges: delicate flesh tones, textures and smooth gradations in deep, saturated colors, and full-tone black and white images. Any serious defect in a color conversion system is unlikely to escape detection here.

Details: Some gamut compression will be required convert these images to these smaller multicolor output spaces: In the first example perceptual rendering intent was used; in the second, relative colorimetric with black point compensation was chosen as the best available option with that system. Black generation (GCR) was adjusted to be as similar as possible in both cases.

System A. This is a conventional ICC-based system, available as a standalone profiling application (to be used in ICC-compliant workflows) and optional color server, which was employed here for the conversion. The underlying measurement data of the profile (for coated packaging board) is plotted in Lab space below. It shows a good balance of shadow, mid-tone, and highlight samples. A modest number of complementary color pairs (C-O, M-G) and extra-color overprints (O-G, V-G, OV, OVG) are present.

The results, shown below, are good. Details and tonal separations are preserved in the deepest and most saturated colors; gray balance is excellent, and no contouring, banding, or posterization is evident. Gradations are smooth.

The OGV separation view below shows a long scale for the OGV channels; they extend deep into colors that are printable with CMYK alone (flesh tones, underside of sunflower).

The graph below, derived from the lower-right image of the jack-o-lantern plant, shows how a range of colors from red-orange to dull green is composed. The extensive interleaving of channels, including small amounts of complementary colors, is a very probable contributor to the visual smoothness of the color transitions.

This system would be a good choice for demanding image reproduction work.

System B. Here we have a non-ICC system sold as an option for a popular workflow suite. It is a bit of a hybrid, its profiling scheme showing echoes of earlier simple spot-color lookup systems: Complimentary colors (C-O, M-G), extra colors (O.G.V) do not overprint, and no build exceeds 4 total colors. Otherwise, its profiles contain an abundance of tint and overprint data, as seen in this Lab plot:

The profile structure is unique, consisting of 4 4-color charts, thus simplifying its design. As we’ll see, simpler is not necessarily better, as when approaching multidimensional problems like image transforms in 7-color space.

The converted test image below predictably shows some differences with Sample A. Color transitions are more abrupt, as shown clearly in the patch chart at the top of the form.

The OGV separation view shows very limited replacement of in-gamut CMY by OGV inks, though this reportedly may be adjusted to some degree in the software. Gradations in these channels are noticeably less differentiated than in System A. The overall appearance of the n-channel separations is more like that of old-fashioned touch colors than fully functioning process channels. The red-green transition contains much less channel interleaving than in Sample A.

Two image comparisons will suffice to show some implications for these different separation schemes. In the lower area of the still life image below we see the dramatic impact of the compacted tonal scale of the violet channel in System B (right). Deep blues are hollowed out and posterized. Faithful rendering of dark, saturated colors is a critical attribute of any good color reproduction system, and this example is decidedly sub-par. System A (left) shows normal results.

The next image detail is a good test for reproduction of subtle color transitions. As seen in the System A image (left), the jack-o-lantern displays a nicely nuanced transition from orange to yellow-green in which a multitude of slightly varying flecks of yellow-green and yellow-orange can be seen. The System B image (right) looks comparatively crude, with a relatively flat, featureless orange abruptly breaking to an undifferentiated yellow-green. (This may be difficult to see in this low-resolution image.) Such defects are not correctable through image retouching, as the separations simply lack the necessary supporting tonal information.

This image also shows an interesting feature of System A. In the transitional color regions we see the presence of the complementary orange-cyan combination:

 These colors, as well as green and magenta, are of course mutually cancelling and therefore commonly regarded as unnecessary in a press separation, a belief so widely accepted that it might be regarded as one of the tenets of extended-gamut printing. As noted earlier, such combinations are excluded from the separation in System B. However, closer study reveals a useful role for these “interstitial” colors in smoothing transitions.

A cautionary note on profile accuracy

An otherwise capable multicolor system may be compromised by the sort of measurement data underlying the profile being used. Here is a Lab plot of a typical 7-color measurement set used by a popular ICC profiling application.

Notice the extreme abundance of dark overprints, far in excess of any possible utility, and the skeletal representation of mid-tones and highlights. This gaping void must be interpolated, literally guessed at by the CMM. The resulting converted images are smooth, but intermediate values are largely fictitious. The less linear the device output is the worse the fiction! There is no known workaround with this application.


Modern multicolor extended-gamut systems must capably convert a variety of elements, including images, vector objects, and spot and process colors, singly and in combination. Systems based on older spot-color matching schemes may have a tougher time converting images compared with systems based on ICC multichannel profiling. Nonetheless, these older systems continue to enjoy popularity in package printing, where the dominance of vector designs gives rise to concerns about printability that may compete with the need for high image fidelity. In the next blog we’ll see how these two schools of thought play out in actual practice, with particular focus on strategies and techniques for converting spot colors. If space allows we’ll also touch upon the arcane subject of EG proofing.

Are you having a multicolor crisis of your own? Are battles over your system configuration leaving you black and blue? Please feel free to post comments and questions here, or send them to

About today's blogger

Mike Strickler is the guy I call on the rare occasions when I want to sully my mind with practical concerns about color management. Mike is a specialist in graphics arts and color management. He is an Idealliance G7 Expert, CMP Color Management Master Trainer and member of PIA and IAPHC. Based in the San Francisco Bay and Los Angeles areas. Specialties: Color management, printing, remote proofing, and photography.

Mike is the principal at MSP Graphic Services.

Tuesday, June 26, 2018

Looking for case studies!

Proof of concept has been established on my ColourSPC project. Over 561K color measurements have been compiled of roughly 3,000 production colors from nine different sources, including data from packaging, newspaper, toner-based, and offset printing, as well as photography, and plastics. The analysis demonstrated that when a color process is in control, the new Zc statistic will follow a specific known distribution.

I am moving to the proof of utility phase, where I hope to show that my new techniques can turn color data into information that is useful for color manufacturers.

I am looking for case studies; people with color manufacturing data and a burning question that they would like answered with that data. Contact me if you want to be part of this exciting new research!

Example questions

Some examples of questions that can be answered with ColourSPC:

    Is this data point an outlier, or just somewhat unusual?
    Was this production run under control?
    What is the major contributor to color variation?
    Did this new piece of equipment or software, or a change in process reduce color variation?
    Can this process reliably meet the color tolerance that my customer wants?


Applications of ColorSPC, Print Properties Council, March 2018