The image importance approach to human vision based image quality characterization

Characterization of image quality is an elusive concept which has been approached in various ways, from quantitative methods based on signal and information theory, to subjective psychovisual or task related assessment. The productive application of these approaches depends closely on the purpose for which the image quality indication is needed, and in many cases the different approaches provide contradictory indications. No satisfactory model has been proposed which caters for a wide range of different purposes, nor for widely differing image properties. This paper describes a flexible approach based on the concept of image importance, which blends several methods together in a form which can be tuned to suit different applications which offers reasonable correlation with subjective results.