The contrast dependence of perceived depth was quantified through a series of depth matching experiments. Perceived depth was found to be a power law function of contrast. In addition, subjects exhibited a large uncrossed depth bias indicating that low contrast test patterns appeared much farther away than high contrast patterns of equal disparity. For disparities in the range of±4.0 arc min, matching disparities for low contrast patterns were shifted in the uncrossed direction by the same amount. In other words, while the magnitude of the uncrossed depth bias is a power law function of contrast, it is constant with respect to disparity. In a second series of experiments, the contrast dependence of stereo increment thresholds was measured. Like perceived depth and stereoacuity, stereo increment thresholds were found to be a power law function of contrast. These results suggest that contrast effects occur at or before the extraction of depth and have implications for the response properties of disparity-selective mechanisms.

Many models and metrics for image quality predict image discriminability, the visibility of the difference between a pair of images. Some image quality applications, such as the quality of imaging radar displays, are concerned with object detection and recognition. Object detection involves looking for one of a large set of object sub-images in a large set of background images and has been approached from this general point of view. We find that discrimination models and metrics can predict the relative detectability of objects in different images, suggesting that these simpler models may be useful in some object detection and recognition applications. Here we compare three alternative measures of image discrimination, a multiple frequency channel model, a single filter model, and RMS error.

To determine whether disparity averaging occurs across spatial scales, a series of depth matching experiments was performed using stimuli composed of the sum of two cosine gratings of different spatial frequency and retinal disparity. Although there were large individual differences in the data, averaging was observed when the relative frequency of the cosines were separated by 3.5 octaves or less and when the relative orientation was less than 30 deg. Beyond these limits, averaging never occurred and the stimulus components were perceived transparently in depth. Furthermore, as the relative contrast of the gratings was varied, perceived depth varied smoothly in the direction of the higher contrast component. This dependence on contrast can be accounted for by a multi-channel model in which separate disparity estimates are generated on each spatial scale and then combined to determine the overall perceived depth of the stimulus.

To determine whether a parabolic template is a good description of the contrast-sensitivity functions (CSF's) exhibited by older adults, the curve-fitting method of Pelli et al. [J. Opt. Soc. Am. A 3(13), P56 (1986)] was applied to contrast-sensitivity data from 100 older subjects (ages 53-85 years). Although the method resulted in reasonable fits for most subjects, closer inspection revealed that this technique may be problematic. A significant number of observers had functions that were nonparabolic, and for many subjects the error tended to be concentrated at the peak of the CSF. In addition, in contrast to the study of Pelli et al., the peak contrast sensitivities of the subjects were only weakly related to Pelli-Robson contrast sensitivity and letter acuity. The data were also fitted with an asymmetric function of variable shape. Whereas this function provided a better fit to the nonparabolic CSF's, it resulted in inferior fits to most of the remaining data. These results demonstrate that the spatial CSF's of older adults cannot be described by a single parametric curve such as a parabola or a function of an exponential and that Pelli-Robson contrast sensitivity and letter acuity are not adequate predictors of their peak contrast sensitivities.

Several stereo-matching algorithms posit that processing on coarse (low spatial frequency) scales constrains processing on fine (high spatial frequency) scales by shifting the disparity range over which fine mechanisms operate. If this were the case, stereo increment thresholds for high frequencies in the presence of low frequencies would be constant, regardless of base disparity. In contrast, we find that in the presence of low spatial frequencies, increment thresholds for high spatial frequencies increase with base disparity, as do increment thresholds for high frequencies presented alone. As a further test of whether information on coarse scales enhances processing on fine scales, diplopia thresholds were measured. In the presence of low frequencies, fusion ranges for high frequencies were severely reduced. These constraints were found to exhibit a power-law dependence on low frequency contrast and to operate over relatively localized areas of the visual field.

It is now generally accepted that depth information is processed by a set of mechanisms, each tuned to a restricted range of disparities. The contrast dependence of stereoacuity argues that these mechanisms are also tuned for contrast. Due to the resulting confounding of contrast and disparity, it should be possible to alter the perceived depth of a test pattern simply by changing its contrast. We tested this hypothesis with a series of depth matching experiments. Over the range of 10–100% contrast, perceived depth varies as a power law function of contrast. In addition, at low contrasts the test pattern always appears to be farther away than the 50% contrast comparison pattern, regardless of whether the test disparity is crossed or uncrossed. In other words, the test appears to have more depth than the comparison in the uncrossed case and less depth (closer to the fronto-parallel plane) than the comparison in the crossed case. Although the test pattern appears to be smaller than the comparison at low contrasts, control experiments indicate that this is not the result of inferring distance from size.

Several stereomatching algorithms posit that processing on coarse (low spatial frequency) scales constrains processing on fine (high spatial frequency) scales by shifting the disparity range over which fine mechanisms operate. To test this hypothesis, we measured stereoacuity and stereo increment thresholds for high spatial frequency test stimuli in the presence of low spatial frequencies. If high spatial frequency mechanisms are always shifted to an optimal range for disparity processing, as proposed in the coarse-to-fine models, stereo increment thresholds for high frequencies in the presence of low frequencies should be constant, regardless of base disparity. In contrast, we find that in the presence of low spatial frequencies, increment thresholds for high spatial frequencies increase with base disparity as do increment thresholds for high frequencies alone. As a further test of whether information on coarse scales enhances processing on fine scales, diplopia thresholds were measured. In the presence of low frequencies, fusion ranges for high frequencies are severely reduced.