A technique for image quality assessment based on a human visual system model

This paper describes an objective quality measure for compressed pictures based on a model of the human visual system (HVS). The structure of natural images and of the coding errors introduced by compression schemes is taken into consideration when determining appropriate HVS parameters. Higher level perceptual factors are also taken into account through the use of Importance Maps. This allows a weighting of the visible errors, since errors occurring in regions of high visual importance have a greater effect on our overall perception of picture quality than those in lower importance regions. The results show the improved predictions achieved by this technique, in particular for images coded with spatially varying quality.