The Image Fidelity Assessor
C.C. Taylor, J.P. Allebach, and Z. Pizlo,
"The Image Fidelity Assessor," Submitted to IEEE Transactions on
Image Processing, February 1999.
Abstract
The IFA is a visual model that
assesses perceived image
fidelity. A set of contrast discrimination experiments that use stimuli
matching the characteristics of receptive field profiles of simple cells
in the visual
cortex serve as the basis for the model. The structure of the model
makes it possible to directly apply the results of the contrast discrimination
experiments to the task of perceived image fidelity assessment.
To evaluate the accuracy of the IFA's predictions,
we propose a psychophysical validation method.
The psychophysical validation method operates
on complex images, retains spatial localization, and seeks to reduce the
effects of response bias. Results from the psychophysical validation
experiments suggest that the IFA can emulate humans in the task of image
fidelity assessment for a wide range of image contents and
distortion types.
Full paper is available in two formats:
t a y l o r@m s o e.e d u