Validation and Verification of a Visual Model for Central and Peripheral Vision
Abstract
Many computational visual models use the contrast sensitivity function (CSF) to represent certain visual characteristics of the observer. In addition, these models are often implemented using a multi-scale, band-limited representation of image contrast. The purpose of the present study was to evaluate a previously described visual model (Peli, JOSA A, 7, 2030, 1990) by comparing the appearance of an image viewed at various distances with simulations of that image corresponding to the same distances generated with the model. Among the unique characteristics of this model are that it applies a threshold (i.e. nonlinear) CSF and a locally normalized, band-limited contrast. Since CSFs can vary substantially depending both on the stimuli and the testing method used to measure them, the model was evaluated using several CSFs. The model was also evaluated for both central images, extending to 20 eccentricity, and peripheral images, extending from 80 to 320 eccentricity. Changes in the images with eccentricity were modeled by a single parameter. For the central (20) stimuli, the CSF obtained with 1-octave Gabor stimuli and a contrast detection task provided better simulations than the other CSFs tested. In addition, data obtained using both lower and higher contrast versions of the same images verified the CSF over a wide range of frequencies and indicated that the model was sensitive to small variations in the chosen CSF. For the peripheral (6.40%320) stimuli, the same 1-octave, detection CSF was found to provide the best simulation. In general, the model suggested by Peli (1990) performed well for both the central and peripheral visual targets, suggesting that the use of a nonlinear CSF and locally normalized contrast are valid.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2000
- Accession Number
- ADP010548
Entities
People
- Eli Peli
- George A. Geri
Organizations
- Schepens Eye Research Institute