Preattentive texture segmentation: the role of line terminations, size, and filter wavelength.

The delta/[symbol: see text] texel pair represents a dilemma in texture discrimination because, despite having the same component orientations, discrimination is still possible (Julesz, 1981), showing a performance asymmetry. Other possible element properties that could influence this task are line terminations, closure, and the size of these elements. We found that line terminators are critical for the task; however, results from double-task experiments indicated that terminator-based discrimination requires the use of attention. When attention is not available for the task, "size" of the elements (with the [symbol: see text] considered slightly larger) seems to be critical for this discrimination and for the asymmetric performance. To generalize the concept of "size" to textures in general, further experiments were performed with textures of different-sized elements. Results showed, as past literature has indicated, that there is a performance asymmetry, with the larger of the elements being more visible when the foreground. This asymmetry was additionally shown to reverse itself (i.e., the smaller element became the more visible) as the scale of the elements increased (while interelement distance remained fixed). A filter analysis was developed in order to measure the apparent size of these elements within textures (texsize), defined as the response weighted average of the filter wavelength, lambda, for a group of elements. The calculation of lambda was attained by introducing a nonlinearity after the second stage of filtering (or spatial averaging of filter responses). This analysis showed high correlation between the texture with the larger lambda and the more visible texture. On the basis of this correlation, a wavelength-dependent noise is proposed, having more internal noise for low-spatial-frequency filters and less for high-spatial-frequency filters.

Dependence on REM sleep of overnight improvement of a perceptual skill.

Several paradigms of perceptual learning suggest that practice can trigger long-term, experience-dependent changes in the adult visual system of humans. As shown here, performance of a basic visual discrimination task improved after a normal night's sleep. Selective disruption of rapid eye movement (REM) sleep resulted in no performance gain during a comparable sleep interval, although non-REM slow-wave sleep disruption did not affect improvement. On the other hand, deprivation of REM sleep had no detrimental effects on the performance of a similar, but previously learned, task. These results indicate that a process of human memory consolidation, active during sleep, is strongly dependent on REM sleep.

Effects of foreground scale in texture discrimination tasks: performance is size, shape, and content specific.

Textural gradients can be defined as differences across space in orientation and spatial frequency content, along with absolute luminance and contrast. In this study, stimuli were created with gradients of these types to see how changing the size and shape of the foreground region affects the psychophysical task. The foreground regions were designed as clusters of target texels alternating with interplaced background texels (of the same cluster size). This design gives rise to a texture square-wave, with texture frequency defined by the distance from the beginning of one target cluster to the next. It was found that for stimuli with vertical and horizontal Gabor patches, the relationship between the global and local orientation of the foreground region is a critical variable, indicating some global-local interaction. When the global orientation of the foreground region is orthogonal to local target texel orientation, visibility is optimal for high texture frequencies, while for parallel arrangements, low texture frequencies are most visible. The latter result was also found to a lesser degree for tasks involving contrast gradients as well as spatial-frequency gradients, but with no effect caused by varying the global orientation. The results indicate the existence of a second-stage filter that integrates (across space) responses of similar first-stage spatial filters, and then sums the resultant activities with those of orthogonal first-stage filters, whose spatial proximity are to the sides of the local orientation. The size of these integrating mechanisms may extend to more than 7 deg, with connections between smoothed activities of filters with orthogonal orientations spanning approximately 1-2 deg.

Spatial variability as a limiting factor in texture-discrimination tasks: implications for performance asymmetries.

Texture-discrimination tasks reveal a pronounced performance asymmetry depending on which texture represents the foreground region (small area) and which represents the ground (large area). This asymmetry implies that some global processes are involved in the segmentation process. We examined this problem within the context of the texture-segmentation algorithm, assuming two filtering stages. The first stage uses spatial frequency and orientation-selective (Gabor) filters, whereas the second stage is formed by low-resolution edge-detection filters. The presence and location of texture borders are indicated by significant responses in the second stage. Spurious texture borders may occur owing to textural local variabilities (such as orientation randomization), which are enhanced by the first stage. We suggest that these spurious borders act as background noise and thus limit performance in texture-discrimination tasks. The noise level depends on which texture occupies the ground in the display. We tested this model on numerous pairs of textures and found remarkably good correlation with human performance. A prediction of the model, namely, that discrimination asymmetry will be reduced when textural elements have identical orientation, was tested psychophysically and confirmed.