This review appeared on pages 19-20 of the Color Business Report, published by Blackstone Research, Vol. 10, Number 7, July 2000.
Psychometric Scaling: Applying Fromal Scaling Techniques to Image Quality Appraisals.
We are all experts on image quality. Just by looking at a print, we can tell pretty quickly whether it is suitable for our purposes. Although the ability to make accurate judgments about image quality is virtually universal, it is not a simple matter to include judgments from human observers in imaging product development programs in an organized and effective way.
Usually, when we collect judgments from observers, we provide a scale for them to use to express their response. "Excellent," "Good," "Fair," and "Poor," is one such scale, called a nominal scale since all of the response choices are named. The techniques of using input from human observers to help determine the scales to be used in such studies is called psychometric scaling, which means, literally, mind measuring.
The scientific underpinnings of psychometric scaling are to be found in disciplines which may be far afield from the usual domain of the materials scientists, mechanical engineers, physicists, and chemists that may make up a printer manufacture's R & D team. Psychologists, psychophysicists, social scientists, and some market researchers use psychometric scaling. Industrial applications for psychometric scaling include the food processing industry, where psychometric scaling supports sensory evaluation.
Image Scientist and Color Business Report's contributing editor Peter Engeldrum has tracked scaling scientists' work down, studied their papers and books, extracted the components that apply to imaging science, and organized the findings into Psychometric Scaling: A Toolkit for Imaging Systems Development. The new book shows how to apply the science of psychometric scaling to imaging product R & D programs. Color Business Report editor Michael Zeis served as editor for Psychometric Scaling.
Could a research project have been more fun than the one conducted by W. J. Dixon and A. M. Mood in the late 1940s? In order to understand the performance of explosive fuses, the Explosive Research Laboratory (Bruceton, PA) scientists dropped a weight from an arbitrary height onto a sample of explosive to see whether the specimen survived or blew up. If there was no explosion, the height was increased, and they tried again. If the sample exploded, the height was decreased and they conducted another trial. This lively example illustrates the development of the "up-and-down method" of reducing error while using the "method of limits" to determine absolute thresholds. (Dixon and Mood published their work, A Method for Obtaining and Analyzing Sensitivity Data, in the Journal of the American Statistical Association in 1948.)
Engeldrum's emphasis in Psychometric Scaling: A Toolkit for Imaging Systems Development is on the practical application of science, rather than on providing a treatise on the theoretical foundations. (A nine-page Bibliography directs curious readers to original papers on the subjects covered and cited.) Although there are plenty of formulas in the text, the text is virtually free of bulky worked examples, which are delivered instead on a CD attached to the inside of the back cover of the book.
Relieved of the burden of writing a science text, Engeldrum has concentrated on summarizing pivotal advances, identifying scaling methods, explaining their mathematical properties, describing data collection and analysis, and identifying the circumstances that make a particular technique or tactic appropriate. It should surprise no one that the most rigid applications of the science are also the most costly. In addition, since human observation is integral to scaling studies, factors such as respondent fatigue must be considered when selecting scaling methods. Therefore, much of Engeldrum's discussion covers the trade-offs one must accept if one uses one method over another.
One does not need a degree in mathematics to benefit from Psychometric Scaling. Although the book was written for physical scientists and engineers, the work can be absorbed by non-scientific program managers whose budgets may fund imaging system development. In addition, those in marketing who sponsor and manage the market research function can use Psychometric Scaling to learn better ways to incorporate image quality appraisal into their studies.
Psychometric Scaling will help manufacturers and software
developers apply more structure to their image quality evaluation
programs. When used in the larger context of Engeldrum's Image
Quality Circle, Psychometric Scaling provides the necessary inputs
to draw relationships between customer perceptions, customer quality
preferences, and technology variables. When such relationships
are made, image quality earns a full-status role in the product
development program. In Engeldrum's Image Quality Circle model,
customer perceptions are known as "nesses," since they
are words like "sharpness" and "darkness"
that end in "ness." Said Engeldrum in the concluding
chapter of Psychometric Scaling, "A propensity exists in
product development environments to go directly for image quality
. Scaling merely for image quality and not the
component "nesses" robs the development team of valuable
information about needed changes in these components. Although
this is not disastrous, it is not resource-efficient,
Psychometric Scaling: A Toolkit for Imaging Systems Development is published by Imcotek Press. ISBN 0-9678706-0-7. The 185-page hardcover book sells for $94.95. Includes Bibliography, Index, Definitions of Symbols, and a CD-ROM containing MathCAD worksheets. Additional information about Psychometric Scaling, and an order form, can be found at here.