ABSTRACT
There are so many attributes that influence image quality that it is clearly impractical to perform factorial experiments that investigate the dependence of quality on combinations of attribute levels. Suppose one were interested in the effect of two attributes on image quality. To span the quality range of practical interest, with acceptable resolution, might require that 8 variations of each attribute be investigated. If 10 scenes were included in the study, and a full factorial design were followed, at 80 quality ruler assessments per one-hour session, it would take 10·82/80=8 hours of evaluation time per observer to rate every sample once. A similar design in which 3 attributes were varied would require 64 hours per observer, and so on.
