A New Measure for the Eeect of Sharpening and Smoothing Lters on Images

It is well-known that the mean squared error (MSE) is an inappropriate measure for the di erence between two images in many applications. For one such an application, edge-preserving smoothing, an alternative was developed which takes both goals into account: the preservation or sharpening of edges and the smoothing of regions. In this paper, tests on human subjects are reported which con rm that the new measures conform reasonably to visual judgement. Next to this, preliminary results are given from experiments in which the preference for sharpening and smoothing is investigated. It is found that images with a relatively low smoothing and a high sharpening are preferred.