Evaluating Color Graphically

From ColorWiki

Jump to: navigation, search
Reserved Article

This page is a
Reserved Article.
For more details see
Reserved ColorWiki Articles


This reserved article originally appeared in CHROMiX ColorNews Issue 7 on April 29, 2003.

Click here to see the original in its original context.
colornews(at)chromix.com to subscribe to the ColorNews newsletter.

by Steve Upton

Evaluating Color Graphically

ICC profiles contain a remarkable amount of information that is often overlooked. The color transformations performed by profiles allow colors to move through our workflows - say from scanner to press - unaltered... or at least that's what we hope. It turns out that each device in your workflow has its own capabilities and that each successive device in your workflow tends to have less and less color reproduction ability. Once we accept this fact of physics we realize that color image reproduction may not be about the most accurate color attainable, but instead is about getting us the most pleasing image we can as the gamut of our image gets squished onto the final output device.

Color as a 3D representation

Whether it's human nature or mathematical necessity, color is typically reduced to three dimensions. XYZ, Lab, LCH, Yxy, Luv and many other 3D color spaces exist to help compare, convert and edit colors.

Example of a 2D graph

The classic shoe-heel-shaped Yxy chromaticity diagram is ubiquitous in color management 101 presentations as it quickly describes how the triangle-shaped gamut of a monitor compares with the blob-shaped gamut of a printing device. It doesn't take long however before we realize that white and black and all of the highlight and shadow colors near them are missing from the 2D graph. Those of us who slave over images spend a lot of time stressing over highlight and shadow reproduction and to have them left out of visualization tools limits our ability to evaluate what's going on in our workflows.

Enter 3D viewing

Plotting the gamut of a printer in 3 dimensions shows us a volume with curves, slopes, points and edges. We can quickly see the brightness and color cast of the paper, how deep the shadows can be and how clear and bright the printer can render saturated colors. Overlay a monitor or scanner profile and limitations in the printing system that were invisible before, become clear. Perhaps clear enough to motivate you to try a different paper type or ink set.

Comparing Gamuts

Several years ago I was struggling in my attempt to have a client's Epson 1270 simulate a Matchprint proof. All the 2D diagrams and manufacturer claims I had seen suggested that the gamut of these new inkjets was big and beautiful and could certainly reproduce anything a press could. If an inkjet printer can reproduce all the colors of a press, then with high-quality profiles you should be able to achieves an acceptable simulation of the press on the inkjet; a proof.

Problem was my test image contained a man's neck tie with a deep dark red that shifted to a grayish brick-red when printed on the 1270. I rebuilt profiles, tried multiple rendering intents and even went so far as to edit both the press and inkjet profiles in an attempt to get them to match. Finally I threw up my hands and I left with my tail between my legs... not a happy day for a color management professional. Some weeks later I wrote a small piece of code that would later evolve into the ColorThink Grapher. It was wire-frame only and low resolution but it allowed me to overlay the gamut volume of the actual Epson 1270 and Matchprint profiles and I was quickly struck with the newly obvious truth. When the gamut comparison of the 1270 and Matchprint are seen in 2D, the 1270 seems to engulf the Matchprint gamut. But bring in 3D graphing and suddenly it becomes apparent that there were whole groups of dark, saturated colors that the Matchprint gamut contained but were not in the gamut of the Epson 1270! A little more graph fiddling and I could see that the reds in my client's image fell in this color contention zone.

So! Finally an answer. No wonder I couldn't get that bloody red to match, the 1270 was simply not able to print it! Did that solve my printing problem? No. The gamut limitations are a real part of the ink and paper combination of the 1270. Still, I would have loved to know that I was dealing with a physical printing limitation before I spent so much time and effort on a fruitless pursuit. The ColorThink Grapher has evolved a long way since those first graphs but the basic technique remains the same. All it takes is a couple of profiles and a good grapher to see whether or not one printer can possibly match another.

Example of image plotting in ColorThink
Device capability rather than profile behavior

Does this mean that once you have the graph showing one gamut engulfed by another that they'll match? No. The 3D profile gamuts in ColorThink are showing more of what the printer is able to reproduce rather than the ability of the profile when used with real images. In order to see the behavior of profiles themselves, actual image data should be used. Choosing a representative data set A good place to start when graphing image colors is an image containing a reasonably wide distribution of colors. Bright and dark, saturated and near-neutral, in gamut and out of gamut (for the printer) colors all help produce a visualization of how your profile affects the colors in your images. Drag an image onto the Grapher in ColorThink and the image's embedded color profile converts the RGB data to Lab.

3D in motion

Graphed in 3D along-side a print profile and it's apparent which colors are printable and which are going to need to undergo "gamut compression" in order to print.

Follow the arrows!

ColorThink can also show how the colors will shift when the print profile is applied. Displayed as vectors, each image color is transformed into an arrow showing the direction and amount of color shift that will occur in your workflow. Finally you can see what is happening to those blues that got clobbered in your great sky shot!

Synthetic data

There are also many different synthetic test data sets available for when you want to isolate the effects to neutrals, certain color ranges, etc. If you can create the file in Photoshop or calculate the color numbers in Excel there is no end to the testing that can be performed. A common use of this technique is to create a small palette of corporate logo colors and then see how well they reproduce (or don't) on different devices: inkjets, monitors, presses, overhead displays, and so forth.

Rendering intents

As you probably learned in your color management 101 class, profiles contain different rendering intents so you can control how the color mapping occurs. Viewing the effects of different rendering intents can help you understand their purpose as well as when one intent is more appropriate than another. You can also compare the rendering intents of profiles from different manufacturers. This level of analysis is the sort of thing that makes same profile manufacturers nervous. It can expose bugs and mis-mappings that will appear unpredictably on your images.

Do these visualization tools solve all the problems you will encounter in your workflow? No, but they will definitely help you understand what is going on. Once you know more about how your color management system works you can start to methodically weed out the problem inks, papers, devices, profiles, applications and so forth. Only by breaking open the "black box" profiles can you begin to see their true capabilities and start to achieve the imaging quality that you are striving for.

Personal tools