A framework for image quality models

Nonlinear image quality models using percepts such as graininess and sharpness as predictors are reviewed. A formalism, the generalized weighted mean hypothesis (GWMH), is suggested as a framework for image quality modeling. This framework is used to develop an image quality model for color business graphics having a prediction error of 2.0 on a 0-to-100 quality scale. Comparison of the GWMH with previous image quality models suggests that it predicts certain model properties.