We present a concept for image quality that is based on a definition of quality in terms of the degree to which something satisfies the requirements imposed on it. An answer to the question of what image quality is must therefore necessarily include answers to these questions: what are images? what are images used for? and what are the requirements which the use of images imposes on them? In this chapter we therefore start by formulating answers to these questions. To this end, we distinguish two main requirements that are imposed upon images. First, the items in the image should be successfully discriminable; and, second, the items in the image should be successfully identifiable. Based on the concept of (partially) flexible metrics presented in Chapter 4, we then formulate algorithms for predicting discriminability, identifiability, and overall performance. To demonstrate the validity of this concept, we compare predictions made with these algorithms with experimentally obtained judgments of human subjects.
|