Anthony Hayes

Contour integration and cortical processing

Our understanding of visual processing in general, and contour integration in particular, has undergone great change over the last 10 years. There is now an accumulation of psychophysical and neurophysiological evidence that the outputs of cells with conjoint orientation preference and spatial position are integrated in the process of explication of rudimentary contours. Recent neuroanatomical and neurophysiological results suggest that this process takes place at the cortical level V1. The code for contour integration may be a temporal one in that it may only manifest itself in the latter part of the spike train as a result of feedback and lateral interactions. Here we review some of the properties of contour integration from a psychophysical perspective and we speculate on their underlying neurophysiological substrate.

Coherent global motion in the absence of coherent velocity signals.

It is widely believed that form and motion are analysed separately in mammalian visual systems. Form is confined within a stream that projects ventrally from V1 to the inferotemporal cortex, and motion within a stream that projects more dorsally, to the posterior parietal cortex [1] [2] [3] [4] [5] [6] [7]. Current descriptions suggest that there is little contact between the two streams until the products of their separate analyses are bound together at a late (and still unidentified) stage in perception [3] [8] [9] [10]. There are, however, indications that form and motion signals may interact [11], and that form signals, streaks derived from motion, may assist in the analysis of its direction [12]. Lennie [13] proposes that all image attributes, form and motion included, remain intimately coupled within the same retinotopic map at all stages of visual analysis. Here we show that form, independent of motion, can give coherence to incoherent motion. Sequences of Glass patterns [14] built to a common global rule are devoid of coherent motion signals, but they produce motion consistent with the global rule for form, not with the random velocity components of the pattern sequence.

The coding of spatial position by the human visual system: Effects of spatial scale and retinal eccentricity

In this study we investigate the nature of the computations that underlie the encoding of spatial position by the human visual system. Specifically, we explore the relationship between alignment accuracy and retinal eccentricity for stimuli where local luminance, local contrast, and orientation cues do not underlie performance. Spatial scale is especially important for such a comparison because of the well documented spatial inhomogeneity of the human visual field. The results suggest that the relationship between spatial localization and eccentricity is invariant with spatial scale if accuracy and eccentricity are expressed in terms of the stimulus envelope size. We show that the photoreceptor disarray does not determine the limit to performance for this task, the limit is post-receptoral and can be modelled in terms of a positional uncertainty within the early filters located before the response envelope has been extracted. This uncertainty varies with eccentricity in a similar way within each spatial array.

Sensitivity to contrast histogram differences in synthetic wavelet-textures.

Recent research on texture synthesis suggests that characterisation of those properties of textures to which human observers are sensitive may be provided by the histograms of the coefficients of a wavelet decomposition. In this study we examined the properties of wavelet histograms that affect texture discrimination by measuring observer sensitivity to differences in the wavelet histograms of synthetic textures. The textures, generated via Gabor micropattern synthesis, were broadband, with amplitude spectra that are characteristic of natural images, i.e. 1/f. We measured texture-difference thresholds for three moments of the wavelet histograms -- variance, skew and kurtosis -- by manipulating the contrast, phase, and density, of the Gabor elements used to construct the textures. Observers discriminated more efficiently between textures that had differences in kurtosis, than between textures that had differences in either variance or skew. Performance was compared to two model observers; one used the pixel-luminance histogram, the other used the histogram of the output of wavelet-filters. The results support the idea that the visual system is relatively sensitive to the kurtosis, or 4th moment, of the wavelet histogram of textures. We argue that higher than 4th-order moments will, in practice, become increasingly difficult for the visual system to represent because the lack of a perfect match between the elements and the receptive fields effectively blurs the response histogram, thereby attenuating higher moments.

Neural recruitment explains “weber's law” of spatial position

We ask whether the well known Webers law between spatial localization and element separation for high contrast, spectrally broad-band stimuli is a consequence of the organization of the early visual filters, or a fundamental constraint on the computation of spatial position by more central mechanisms. We address this question by identifying the individual contributions of mechanisms tuned to different ranges of spatial frequencies and contrast. We measure spatial-alignment and bisection error as a function of element separation at each of a number of spatial scales, using spectrally narrow-band stimuli of fixed supra-threshold contrast. We show that stimuli which minimize the extent of neural recruitment across different spatial channels before the site of extraction of the local contrast energy (and to a lesser extent across different contrast channels) do not exhibit Webers law for either alignment or bisection. We present evidence that Webers law for localization with increasing separation, found for stimuli of high contrast and broad-band spatial frequency content, is a consequence of the successive disengagement of unitary neural mechanisms, each of which has different spatial and contrast properties, and none of which individually exhibits Webers law for spatial position.

The role of luminance contrast in the detection of global structure in static and dynamic, same- and opposite-polarity, Glass patterns

Perception of global structure conveyed in static Glass patterns is difficult, though not impossible, when the constituent dipoles are formed by partnering opposite polarity dots. We investigate whether the addition of motion signals to opposite-polarity Glass patterns can act to restore the perception of global structure. The stimuli were concentric Glass patterns consisting of 200 dipoles concentrically orientated, or oriented at random orientations, placed on a grey background. For each dipole, one luminance-increment dot (Weber contrast of 1) was paired with another dot set to a contrast ranging between luminance increment and luminance decrement (i.e., a Weber contrast range of approximately -1 to 1). Dipoles were either stationary (Experiment 1), or randomly re-positioned at 17Hz (Experiment 2), on each frame transition. A two-interval forced-choice paradigm, in conjunction with an adaptive staircase, was used to obtain Glass-pattern detection thresholds. The task required observers to identify the interval that contained concentric Glass structure; the other interval contained randomly orientated dipoles. Generally, lower global form thresholds were observed for dynamic and same-polarity Glass patterns than for static and opposite-polarity Glass patterns. In particular, for dynamic presentations improvement in sensitivity was more evident for opposite-polarity than for same-polarity Glass patterns. These findings suggest that motion plays an important role in the detection of global structure in dynamic Glass patterns.

On the role of second-order signals in the perceived direction of motion of type II plaid patterns

Second-order Type I and Type II plaids were constructed by combining two random-dot gratings. Each component consisted of a dynamic random-dot field, the contrast of which was modulated by a drifting sinusoidal grating. Orienting the two components suitably and interleaving at 120 Hz allowed us to produce a two-dimensional plaid pattern made from one-dimensional second-order components. The perceived direction of motion of both Type I and Type II plaids was measured as a function of stimulus duration. Type I plaids had a perceived direction close to the intersection of constraints/vector sum solution (which only coincide for these patterns) at all durations. Type II plaids had a perceived direction that moved away from the vector sum and toward the intersection of constraints solution as the duration of presentation increased. These results are similar in form to those found for plaids made from first-order (luminance-defined) components [Yo & Wilson (1992), Vision Research, 32, 135-147]. This suggests that a delay which operates specifically on second-order signals cannot be the sole cause for the change in perceived direction of Type II plaids made from first-order components [Wilson, Ferrera & Yo (1992), Visual Neuroscience, 9, 79-97].

Moving Glass Patterns: Asymmetric Interaction between Motion and form

The perceived motion direction of a moving Glass pattern is influenced by the orientation of the dot pairs (dipoles) that generate the pattern (Krekelberg et al, 2003 Nature 424 674–677; Ross, 2004 Vision Research 44 441–448). Here, we investigate how the motion vector and the dipole orientation of moving Glass patterns influence the perceived orientation of each. We employed 1 s movie presentations of sequences of linear Glass patterns, each consisting of 200 dot pairs. Signal pairs, aligned in a common orientation, moved in a common direction. The observers task was to indicate either the perceived direction of motion, or the perceived dipole orientation of Glass patterns that consisted of either same-polarity dipoles, or opposite-polarity dipoles. Perceived orientation or motion direction was measured as a function of the angular difference between the orientation and the motion direction of the dipoles. We found that the apparent global direction of motion was attracted by approximately 4° towards the dipole orientation for small (15°, 23°) angular differences between dipole motion-direction and dipole orientation, regardless of dipole polarity. However, under the same stimulus conditions, the apparent global orientation was much less affected by the direction of motion, suggesting that motion and form interact asymmetrically. Global form influences global motion-direction perception more powerfully than global motion influences global form perception.

An experimental comparison of viewpoint-specific and viewpoint-independent models of object representation.

Four experiments that investigate the cognitive representation of objects in human observers are reported. Two broad classes of theory were examined: viewpoint-specific and viewpoint-independent models. The former postulate that the data structures underpinning object recognition correspond to discrete views and require additional processing to access them from unfamiliar viewpoints. The latter postulate data structures that are independent of any particular viewpoint and can be directly accessed from a wide range of viewpoints. Two experimental tasks were used: a sequential matching paradigm and a cognitive learning paradigm. Findings favour viewpoint-specific models over viewpoint-independent models.

Distortion in perceived image size accompanies flash lag in depth.

The flash-lag effect -- a misperception that a flashed object appears to lag behind a moving object despite their physical alignment -- has mainly been investigated as a spatiotemporal offset. Here, we report that the flash-lag-in-depth effect is accompanied by an illusory change in the apparent size of the flashed object. We found a strong flash-lag-in-depth effect with a dot-defined square, whose motion in depth was signaled by changing retinal disparity (stereomotion), and a Gaussian blob that was flashed in the center of the square. Using the same stimulus, observers matched the apparent size of the flashed blob with a reference blob when the square moved with approaching or receding motion. Approaching motion of the square resulted in a reduction in the apparent size of the flashed blob, and an apparent enlargement of the flashed blob was induced by receding motion of the square. Additionally, this size effect substantially diminished, or was eliminated, when looming (change of size) instead of stereomotion was used to cue motion in depth of the square. The flashed-object size change that is induced by the moving square is not explained by simple predictions from projective geometry.