ABSTRACT
Chapter 11 discussed the challenge of predicting multivariate image quality from a knowledge of the impact of each of the contributing attributes in isolation. Pairs of attributes were classified as perceptually dependent or independent based on whether the level of one attribute affected the perception of the level of another attribute. It was stated that in practice, most image quality attributes, if defined carefully, are approximately perceptually independent at most practical levels. In this chapter, two examples of interactions between perceptually dependent attributes are considered; these are: (1) the impact of isotropic noise on streaking, a weak interaction discussed in Sect. 17.2; and (2) the effect of pre-quantization isotropic noise on contouring, a strong interaction discussed in Sect. 17.3. In both cases, the objective metrics of the affected attributes can be extended so that the perceived level of the affected attribute in the presence of the affecting attribute may be predicted. This observation leads to some interesting deliberations in Sect. 17.4, which suggest that the multivariate formalism may be even more generally applicable than initially proposed.
