Biomedical subjects
Publications and source records attributed to B Julesz.
Recent discoveries of nonlinear perceptual analyzers in effortless texture discrimination cast serious doubt on the usefulness of Fourier image decompositions to describe suprathreshold visual-texture perception. We now explain the meaning of these results.
It is known that the sum of a random-dot array with vertical bilateral symmetry and one with horizontal bilateral symmetry appears as a random array. Here we show that if the vertically and horizontally symmetrical arrays are spatially filtered, so that their respective spectra are 2 octaves apart, then their superposition does not appear random, but both symmetries can be simultaneously perceived. The low-band array has a stronger perceptual weight than the high-band array. These demonstrations give further evidence that frequency channels are before symmetry perception.
By defining texture as a global feature attained by integration over the image domain, we show that texture discrimination can be predicted for a special class of visual textures (composed of paired dots) as a function of such global features. We derive a psychophysical law based on these global features and show that differences in the variance of orientation, but not of dipole length, result in texture discrimination.
We found a new class of two-dimensional random textures with identical third-order statistics that can be effortlessly discriminated. Discrimination is based on local "granularity" differences between these iso-trigon texture pairs. This is the more surprising since it is commonly assumed that texture granularity (grain) is determined by the power spectrum which, in turn, can be obtained from the second-order statistics. Because textures with identical third-order statistics must have identical second-order statistics (i.e., identical power spectra), visible texture granularity is not controlled by power spectra, and not even by third-order statistics.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
The latency time of tracking dynamic random-dot stereograms can be shortened by as much as 100 ms when monocular cues are added by introducing a difference in dot density between target and surround. It has been tacitly assumed that perception time will be reduced only if the added monocular cues are above the detection threshold for each eye. However, the experiments reported here clearly show that stereoscopic performance as measured by an eye tracking task can be greatly enhanced by added monocular cues that cannot be detected. Observers were instructed to track a suddenly displaced vertical bar (portrayed as a dynamic random-dot stereogram) while their eye movements were recorded by EOG. The bar had either a given binocular disparity or zero binocular disparity with respect to its surround. For the target with a disparity (in a wide range), the latency time of tracking decreased by more than 30 ms (10%) as density difference increased from 0 to 4%, whereas in the control conditions with no stereoscopic cues (zero disparity) subjects were unable to track the bar at all within that range of density difference. Thus stereopsis is greatly aided by minimal monocular cues that by themselves elude monocular detection.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Dynamic random-dot stereograms devoid of all monocular depth cues were used to measure the limits of temporal and spatial resolution in the center of the visual field. The temporal durations for detecting a small, briefly presented test square of different binocular disparity than the surround varied as a function of its location and binocular disparity. The test squares presented in the upper hemifield were detectable at consistently shorter durations than those presented in the lower hemifield for a surround disparity which was uncrossed relative to the fixation marker. For crossed surround disparity this preference reversed, resulting in a superiority of the lower hemifield. The anisotropy diminished for zero surround disparity. No such anisotropy was found when left and right visual hemifields were compared. It was also shown that this upper-lower temporal anisotropy (and left-right isotropy) is paralleled by a similar disparity-dependent upper-lower anistropy (and left-right isotropy) in spatial resolution. Introduction of monocular clues into the stereograms tended to eliminate the anisotropies. This implies that the anisotropies reflect the spatiotemporal properties and distribution of binocular disparity detectors in the human cortex and result in a tilted surface that pivots around the horizontal midline in the space of binocular depth perception.
With the use of dynamic random-dot sterograms (which are devoid of all monocular depth cues), the temporal duration for detecting a small, briefly presented test square of different depth than the surround varied as a function of its location in the central portion of the visual field. Test squares presented in the upper hemifield were detectable at consistently shorter durations than those in the lower hemifield when the fixation marker was in front of the surround, and vice versa when the marker was behind. No such anisotropy was found for left and right hemifield. Esploratory studies suggested a similar up-down anisotropy and left-right isotropy in spatial resolution. Thus, the upper hemifield representation at the cortex shows a general superiority over the lower one for vinocular detectors tuned to uncrossed disparitites, and the lower hemifield shows superiority for those tuned to crossed disparities.
Explore the source record for details and available documents.
Explore the source record for details and available documents.