ABSTRACT
The previous chapter discussed the design of objective metrics correlating with individual color and tone reproduction attributes. This chapter will demonstrate that the ability to predict color and tone reproduction quality can be substantially extended by transformations applied in perceptual space. The reader will recall that an operation is said to be carried out in perceptual space if its inputs and outputs are quality changes expressed in JNDs. Examples of operations carried out in perceptual space include the multivariate formalism (Ch. 11) and the weighting schemes applied to attributes that vary across an image (Ch. 15). These contrast with the extensions of metrics, such as the incorporation of terms reflecting attribute interactions (Ch. 17), which are accomplished in objective space. It is not always obvious whether an observed effect should be modeled in objective space, by generalizing a metric, or in perceptual space, through some quality transformation. A rough guideline might be that lower level visual processes such as masking are often amenable to treatment in objective space, whereas higher level cognitive effects are better emulated in perceptual space.
