Discrimination of wide-field images as a test of a peripheral-vision model.

In order to test a model of peripheral vision, various contrast sensitivity functions (CSF's) and fundamental eccentricity constants (FEC's) [see J. Opt. Soc. Am. A 8, 1762 (1991)] were applied to real-world, wide-field (6.4 degrees-32 degrees eccentricity) images. The FEC is used to model the change in contrast sensitivity as a function of retinal eccentricity. The processed test images were tested perceptually by determining the threshold FEC for which the observers could discriminate the test images from the original image. It was expected that higher CSF sensitivity would be associated with higher FEC's; and in fact, for images processed with low-pass (variable-window stimuli) CSF's, the threshold FEC's were larger for the higher-sensitivity (pattern-detection) CSF than for the lower-sensitivity (orientation detection) CSF. When two higher-sensitivity CSF's were compared, the bandpass (constant-window stimuli) CSF resulted in essentially the same FEC threshold as did the low-pass (variable-window stimuli) CSF. The fact that the FEC compensated for complex differences in the form of the CSF suggested that the discrimination task was mediated by a limited range of spatial frequencies over which the two CSF's were similar. Image contrast was then varied in order to extend the range of spatial frequencies tested. The FEC's estimated with the lower-contrast test images were unchanged for test images obtained with the high-sensitivity, bandpass CSF but increased for test images obtained with the high-sensitivity, low-pass CSF. These results suggest that peripheral contrast sensitivity as used in the present discrimination task is based on a high-sensitivity, bandpass CSF. The peripheral-vision model validated by the present analysis has practical applications in the evaluation of wide-field simulator images as well as area-of-interest or other foveating systems.