Richard Harvey, Stephen King, Richard Aldridge
and J. Andrew Bangham
School of Information Systems
University of East Anglia, Norwich NR4 7TJ
Tel: +44 1603 593257, Fax: +44 1603 593345.
Interpolating images is a very common task but many interpolation filters are designed using error metrics that, in comparison to the known human vision system, are fairly crude. In this paper a model of the human vision system is used to produce a quality score that is useful for comparing the errors due to image resampling. It is tested with a number of image resamplers. We conclude that such a model produces a measure of image quality that is superior to signal-to-error ratio, but that further refinements are necessary.
Stephen King ESE PG