PROGRAMME - SPRING TERM 1995
venue: Department of Psychology, University of Reading
A new theory of visual object recognition that is based on multi-dimensional interpolation between stored templates requires fast, stimulus specific learning in the visual cortex. And indeed, performance in a number of perceptual tasks improves as a result of practice. We distinguish between two phases of learning a vernier acuity task, a fast one that takes place within less than 20 minutes and a slow phase that continues over 10 hours of training and probably beyond. The improvement is specific for relatively 'simple' features, such as the orientation of the stimulus presented during training, for the position in the visual field, and partly specific for the eye through that learning occurred. Some of these results are simulated by means of a computer model that relies on object recognition by multidimensional interpolation between stored templates.
Orientation specificity of learning is also found in a jump-displacement task. Parallel to the improvement in performance, cortical potentials evoked by the jump-displacement tend to decrease in latency and to increase in amplitude as a result of training. The distribution of potentials over the brain changes significantly as a result of repeated exposure to the same stimulus. The results of both psychophysical and electrophysiological experiments indicate that some form of perceptual learning might occur very early during cortical information processing. The hypothesis that vernier breaks are detected 'early' during pattern recognition is supported by the fact that reaction times for the detection of verniers depend hardly at all on the number of stimuli presented simultaneously. Hence, vernier breaks can be detected in parallel at different locations in the visual field, indicating that deviation from straightness is an elementary feature for visual object recognition in humans that is detected at an early stage of pattern recognition.
venue: Department of Psychology, University of Surrey
Three experiments examined direction discrimination in temporally interleaved random dot patterns. The stimulus consisted of two or more uncorrelated random patterns presented in a repeating temporal sequence, so that each pattern appeared only once every n frames separated by uncorrelated patterns. Each pattern shifted either leftward or rightward at each re-appearance (all patterns shifting in the same direction in any one presentation). Subjects could specify shift direction correctly even when eight different patterns were interleaved, provided that the duration of each frame was brief. An explanation based on responses in first-order motion energy detectors tuned to low spatiotemporal frequencies (effectively summating the interleaved patterns over time) was tested using a stimulus in which each pattern inverted in contrast mid-way through each frame.
Contrary to predictions based on temporal summation, performance with contrast-inverting patterns was only slightly lower than with non-inverting patterns. An alternative explanation was examined, based on responses in second-order motion detectors which rectify image contrast before extracting motion energy. Computed responses from such detectors successfully predicted psychophysical performance with interleaved random patterns. Implications for models of motion analysis were discussed.
The paper this seminar was based on was co-authored by H.Tunley.
venue: Dept. of Psychology, Royal Holloway, Univ. of London
Vision uses two broadly different strategies in the representation of motion from image sequences: direction-selective filtering, and feature or 'token' matching across space-time. I shall review recent evidence that confirms this broad distinction, and reveals more about the spatio-temporal and binocular properties of the two processes. I shall also discuss the relation between direction selectivity and spatio-temporal gradient models, in which velocity can be extracted from spatial and temporal derivatives without the explicit use of directional filters. Experiments on contrast detection at threshold imply that moving and flickering gratings, and even single flashes of a grating, are detected by direction-selective filters. In two-flash experiments the variation of grating threshold with spatial phase offset and time delay between the flashes shows a good fit to the predictions of the Adelson-Bergen motion energy model. Above threshold, experiments on perceived direction for gratings that are both displaced and contrast-modulated over time suggest a modification to the final stages of the energy model. If we denote the two directional energy response values as L and R, then instead of computing direction from motion opponency (L-R) our experiments suggest that direction may be obtained from "motion contrast" (L-R)/(L+R).
We explored the spatial filter properties of motion perception with a variety of spatio-temporal sequences, in which local phase shifted to the left and to the right in alternate strips or columns of the pattern. We assume that perceived motion breaks down when two strips containing opposite motion occupy the receptive field of the mechanism. On this basis the receptive fields for motion are quite small, and spatially very broad-band, confirming by a very different technique the conclusions of Anderson & Burr (1). For example, at 1 c/deg the s.d. of the receptive field envelope is only 6 min arc. Such broadband filters offer a direct low-level basis for the coherent motion seen in moving plaids.
A contrast-modulated (CM) pattern is formed when a modulating or envelope function imposes local contrast variations on a higher-frequency carrier. Motion may be seen when the envelope drifts across a static carrier and this has been attributed to a 'second-order' pathway for motion. However, an early compressive response to luminance (e.g. in the receptors) would introduce a distortion product at the modulating frequency. We used a nulling method to measure the distortion product, and then asked whether this accounted for 'second-order' motion. Distortion amplitude increased as the square of modulation contrast, as predicted by the compressive transducer. It also increased with modulation drift rate, implying that the transducer is not static. Thus early compressive non-linearity induces first-order artefacts into second-order stimuli. Nevertheless this does not account for second-order motion, since in a second study perceived motion of second-order sequences (CM in every frame) could in general not be nulled by adding luminance modulated (LM) components. We conclude that two pathways for motion do exist.
References
1. Anderson, S. J. and Burr, D. C. (1991) Spatial summation properties of directionally selective mechanisms in human vision. J. Opt. Soc. Am. A., 8, 1330-1339.
venue: Department of Human Sciences, Brunel University