IEEE/EMBS Summer School'96
Objective Evaluation of Image Quality

Objective Evaluation of Image Quality

Harrison H. Barrett

University of Arizona

Tucson, U.S.A.

---

In art and advertising, image quality can be entirely subjective; beauty is truly in the eye of the beholder. When images are produced for scientific and medical purposes, however, objective definitions of iamge quality are needed. A great deal of research goes into development of better imaging techniques, but we cannot know if this research is successful unless we give a rigorous definition of image quality and develop a methodology for measuring and optimizing it.

A rigorous definition of image quality relates it to how well desired information can be extracted from the image. In other words, image quality is measured by the performance of some "observer" on some specific task. The observer can be a human, such as a physician trying to make a diagnosis, or it can be a computer algorithm. The tasks can be divided generically into classification or estimation tasks. In medical applications, an example of aclassification task would be lesion detection, while an estimation task might be determination of the volume of blood expelled from the heart on each beat. For classification tasks performed by a human observer, psychophysical studies and ROC (receiver operating characteristic) analysis provide a robust, objective measure of image quality. Such studies are, however, time consuming and imprecise, and they do not provide an easy way to see how image quality varies with various parameters of the imaging system or processing algorithm. For these reasons, there is considerable interest, especially in the radio logical literature, in mathematical model observers that can substitute for the human observer.

In this talk we survey current efforts to develop model observers and present several case studies showing how they can be used to optimize imaging systems and image-processing algorithms.

Back to the Summer School home page.