Several color image fidelity metrics are evaluated by comparing the me
tric predictions to empirical measurements. Subjects examined image pa
irs consisting of an original and a reproduction. They marked location
s on the reproduction that differed detectably from the original. We r
efer to the distribution of error marks by the subjects as image disto
rtion maps. The empirically obtained image distortion maps are compare
d to the predicited visible difference calculated using (1) the widely
used root mean square error (point-by-point RMS) computed in uncalibr
ated RGB values, (2) the point-by-point CIELAB Delta E-94 values (CIE,
1994) and (3) S-CIELAB Delta E-94, a spatial extension of CIELAB Delt
a E metric. The uncalibrated RMS metric did not predict the perceptual
image distortion data well. The point-by-point CIELAB Delta E-94 metr
ic provided better predictions, and the S-CIELAB metric, which incorpo
rated the spatial color sensitivity of the eye, gave the most accurate
predictions. None of the metrics provided an excellent fit to the dat
a. Image areas with poor predictions were concentrated in regions cont
aining large negative local contrast. When these areas were excluded f
rom our data analysis, both S-CIELAB and CIELAB predictions had much b
etter agreement with the perceptual data. This suggests that the next
step in improving color image fidelity metrics is to redefine color di
fference formula such as CIELAB Delta E-94 in terms of local contrast.
(C) 1998 Elsevier Science B.V. All rights reserved.