[10.24.2001]
Interview with Robin Myers: Part 2
Robin: The widespread availability of tools for profile generation, profile editing, and the continued adoption of ColorSync and ICC color management in applications is great! It means easier production of color for users and that benefits everyone. It is what Gary Starkweather and I wanted when I started working on the prototype ColorSync 1. The biggest shortcoming of the current ICC approach is the almost complete reliance on L*a*b* as the device independent Profile Connection Space. L*a*b* was never intended for color matching calculations and has some extremely bad features that make life difficult. Almost everyone I know has had problems with skies in images turning purple. This is solely due to a quirk of L*a*b* space that some blues change hue and lightness as the saturation increases! If the profile generation software does not automatically correct for this, the profile must be tuned to reduce this problem. There are other color calculation anomalies with L*a*b* space; that is why it was intended only for use in comparing color differences.
Additionally, an exact colorimetric match for in-gamut colors is very difficult to achieve with an ICC profile. In what I think was an attempt to reduce the profile size, the ICC only includes a relative colorimetric table and uses a mathematical transformation to create an absolute colorimetric result. However, my tests have shown that this approach results in reproduction errors. Use of XYZ as a Profile Connection Space instead of L*a*b* and the inclusion of an absolute colorimetric table would improve the quality of the profiles. To be fair, the ICC does allow for the use of XYZ as a Profile Connection Space, but almost all the profile generation and editing tools are L*a*b* only. In fact, if you send an ICC profile with XYZ as the Profile Connection Space into ColorSync 2 or 3, it will convert the tables to L*a*b* before it will do any calculations with them.
As a little historical note, the color matching engine of ColorSync 1 used XYZ as its Profile Connection Space and for absolute and relative colorimetric matching it was more accurate than the current ICC approach.
Oh, there is one more objection I have to the current ICC design and that relates to camera profiling. The ICC has dictated that only the relative colorimetric intent is to be used for scanners or digital cameras and there is a very common professional need for an absolute colorimetric rendering intent. As noted above, producing absolute colorimetric data from transformed relative colorimetric data is not accurate enough for some applications.
Jeff: Share with us a little more about your background, where you've been, what you've done. For example, how did you develop such an interest in imaging, and in color science in particular?
Robin: My formal training is in Chemistry but I've always had an interest in many areas of science and technology. I used some college summer jobs as a chemist to earn the money to purchase one of the first microprocessor computers and used the knowledge gained in working with this computer to get a job as a programmer at Lawrence Livermore National Laboratory. There I worked with computer graphics, real-time programming, database design, simulation and other applications.
This experience and my hobby of Cibachrome color photographic printing enabled me to get a job at Versatec working on the development of software for the first electrostatic color printer. This was a wide format color raster printer a decade before wide format inkjet printers existed. At Versatec I worked on graphics systems design and implementation, color matching and represented the company on some ANSI standards committees. The research on color lead to my first patent for a color matching algorithm and this helped me get a job at Apple Computer.
At Apple, I worked on many topics in imaging and other fields. Some of the projects involved color matching, image processing, printing, typography, digital photography, digital stereo photography, end user programming, optical character recognition and others. I've received 3 more patents on color matching algorithms from my Apple work. Looking at my career and personal interests, there has always been a common thread of imaging running throughout. My interest in color science really began when I was working at Versatec, trying to solve some color problems with the electrostatic printer. Since then I have been voraciously reading everything I can get on the subject and its application to everyday color issues.
Jeff: Now, tell us about the present. You're a programmer, color scientist, photographer, and color management consultant. What are your current projects?
Robin: As you know, I have currently two jobs. I work at Better Light, a high-resolution digital camera manufacturer, doing research, tech support and some marketing. I also represent Better Light on a standards committee developing a new color target for digital camera calibration and profiling. My second job is for my own company where I am a color management consultant, photographer and imaging consultant. In addition, I have spent the last two years developing a program for spectral databasing, computation and analysis, called SpectraShop™.
Jeff: How do you feel your software differs from what's already being offered?
^ return to top
« return to interviews list
« return to resources list