|
Promoting vision research and its applications |
|
||||||||
|
AVA 2000 AVA Annual Meeting - Motion and Imaging 15 March 2000 Meeting Abstracts GEOFFREY J. BURTON MEMORIAL LECTURE Search for Contrast-Defined Motion H.A. Allen*, A.M. Derrington** (School of Psychology, University of Nottingham, University Park, Nottingham NG7 2RD; e-mail: *lpxhaa@psychology.nottingham.ac.uk, **andrew.derrington@nottingham.ac.uk) We investigated whether observers could process multiple patches of contrast-defined motion simultaneously. We measured the duration required by observers to discriminate between two different four patch patterns of either luminance - or contrast-defined motion. In the first experiment, motion within the patches was consistent with global rotation. In one interval of temporal 2AFC task, all four patches contained motion consistent with anticlockwise rotation and in the other interval, 1,2 or 4 patches contained inconsistent motion (opposite direction). In the second experiment, all four patches contained leftwards motion in one interval and 1,2 or 4 contained inconsistent motion in the other interval. In both experiments, observers indicated which interval contained inconsistent motion. The two experiments produced similar results. For luminance-defined patterns, we found that the number of patches of inconsistent motion had little effect on the presentation duration required by observers to discriminate between the patterns. For contrast-defined motion, as the number of patches of inconsistent motion increased there was a rapid decrease in the duration required. We estimated the minimum duration that would be required for observers to check each patch in turn for both experiments. For luminance-defined patterns, observers performed the tasks much quicker than this estimate. For contrast-defined patterns, they required presentation durations far longer than our estimate. We conclude that multiple patches of contrast-defined motion cannot be processed simultaneously.
How do we process biological motion? Ian M. Thornton Max-Plank-Institute for Biological Cybernetics, Tübingen, Germany Johansson’s point-light walker figures remain one of the most powerful and convincing examples of the role that motion can play in the perception of form. How exactly does our visual system take a few isolated points of light and provide us with such compelling impressions of human action? Here I present data from a series of studies demonstrating that biological motion processing can come to rely almost exclusively on EITHER bottom-up or top-down processing mechanisms. Using a dual-task paradigm I will show that when attention is allocated to a demanding secondary task (change detection), direction discrimination performance drops from near ceiling levels (85% plus), when walkers are masked by randomly moving elements, to complete chance (50%), when more complex, scrambled walker masks are used. This dissociation under dual-task conditions adds further weight to the suggestion that the high, single-task performance commonly reported with both random and scrambled masks is achieved via very different, yet equally effective, processing mechanisms. A second series of studies will further explore the role of attention during biological motion processing using standard visual search techniques. In conclusion, I will argue that the effective use of both high and low-level integration strategies is highly adaptive, given the ecological significance of human and animal action, and may well be at the core of what appears to make biological motion "special".
Motion Opponency in Motion Detection. Edward Powell & Mark Georgeson School of Psychology, University of Birmingham Several theoretical models of motion detection (e.g. Adelson & Bergen, 1985; Simoncelli & Heeger, 1998) incorporate an opponent stage of analysis, in which motion direction is coded by the difference in output between opposite (e.g. leftward and rightward) detectors. Psychophysical evidence for motion opponency, however, has been weak or disputed. We used a two-alternative forced-choice staircase procedure to measure contrast thresholds for moving sinewave gratings (0.5 c/deg, 8 or 20 Hz drift rate) superimposed on a counterphase flickering mask of the same spatial and temporal frequency. Low contrast maskers made movement detection easier (facilitation) while higher contrast masks made it more difficult (masking), resulting in a characteristic dipper-shaped threshold function. The slope of the masking function was close to 1 (Weber's Law). Importantly, sensitivity to motion depended on the difference between opposite motion signals. Sensitivity doubled when the test signal (an increase in rightward energy) also had a decrease in leftward energy (or vice-versa). A motion energy model that incorporates subtractive motion opponency and a divisive gain control - with only 2 free parameters - successfully accounts for both unmasked detection and masked discrimination performance.
A model of motion adaptation and motion after-effects based upon principal component regression Keith Langley Department of Psychology, University College London. A gradient-based model to help explain effects of adaptation to moving signals is compared with established energy models of motion detection. The novel feature of the proposed model is that measurements of the image signals are assumed to be subject to error in both dimensions of space and time. This assumption, when combined with a maximising rather than nulling strategy for adaptive effects (see Langley, AVA Meeting Aston, 1999) necessarily constrains models of motion perception to be based upon principal component regression rather than linear regression. It is shown, that response suppression of (model) complex cell neurons that input into the model may account for (i) significant increases in perceived speed after adaptation to static patterns and testing with slowly moving patterns, (ii) increases in perceived speed after adaptation to patterns moving at a medium speed and testing at high speed, and (iii) decreases in perceived speed in the opponent direction to a quickly moving adapting signal. Neither of predictions (ii) or (iii) are supported by established energy-based accounts of motion detection. It is concluded through comparisons of the proposed model’s speed transfer function with existing psychophysical data that the visual system processes motion signals with the tacit assumption that image measurements are subject to error in both space and time.
Complex Motion Detection in Human Vision T. S. Meese Neurosciences Research Institute, Aston University, Aston Triangle, Birmingham B47ET, UK; Fax: +44 121 333 4220. Four 2IFC summation experiments were performed to explore properties of complex motion detectors in human vision. Coherence thresholds (% of signal dots) were measured for detecting one or two signals (from: expansion, rotation, deformation & translation) in noise for random-dot kinematograms (eg. Harris & Meese, Perception, 25 supp, 129). Stimuli contained up to 440 randomly positioned dots and were viewed through a luminance damped annular window (5 deg and 0.5 deg diameters). Noise dots (random directions) were always contained within the entire display. In Experiment 1, signal dots were contained in either (a) the upper quadrant alone or (b) the entire display. For (b), thresholds fell within the range of perfect linear summation and quadratic summation amongst the four quadrants, depending upon stimulus condition and observer. These results are generally consistent with motion detectors that sum linearly across multiple motion directions; less than perfect linear summation is shown to be consistent with nonlinear summation (eg. probability summation) between detectors that are suboptimal for the stimuli (eg. one-dimensional ‘shear’ and ‘compression’ mechanisms; see Meese & Harris, 1997, Vision Research, 37, 2369-2379). In Experiment 2, summation was typically less than quadratic for pairs of orthogonal stimuli (eg. expansion and rotation), suggesting probability summation between independent detectors. In Experiment 3, direction (d) bandwidths for summation were found to be broad, [approximating cosine(d)], for all stimulus types, suggesting detectors with broad stimulus selectivity. In Experiment 4, substantial summation was found for pairs of stimuli with opposite directions of motion (eg. expansion and contraction) for all stimulus types, except deformation, where performance degraded. This last result suggests an inhibitory input for deformation ‘detectors’, but not for expansion, rotation and translation ‘detectors’.
Object-based selection operating on a spatial representation made salient by dimensional segmentation mechanisms R.B. O’Grady1, H.J. Muller1,2 1)School of Psychology, Birkebeck College, University of London, Malet Street, London WC1E 7HX, 2)Institut fur Allgemeine Psychologie, Universitat Leipzig, Seeburgstrasse 14-20, D-04103 Leipzig, Germany. Three experiments re-investigated selective attention in the ‘ring-cueing’ paradigm of Egly and Homa [1984; Journal of Experimental Psychology: Human Perception and Performance, 10, 778-793]. Observers were cued to attend to one of three concentric rings of radius 1°, 2° or 3°, and their target-detection accuracy on cued and uncued rings was measured (using sensitivity parameter Az as dependent variable). Experiment 1, which used a central colour cue to indicate a like-coloured ring, replicated the ring-cueing effects reported by Egly and Homa. Experiments 2 and 3 examined whether these effects were produced by observers exploiting secondary-depth cues in the display (i.e., although drawn on a flat plane, the display might have suggested a tunnel receding in depth). With colour cues, the availability of secondary-depth information relative to the deliberate disruption of secondary-depth information, had no significant influence on the ring-cueing effects. However, making the rings monochrome and using a central size cue (central symbol whose radius: large, intermediate or small, indicated which ring: outer, middle or inner, was most likely to contain the target) significantly reduced the ring-cueing effects when the depth information was disrupted. The results suggest that selection was object-based, operating on a spatial ‘grouped-array’ representation of the cued ring made salient by colour or depth-based segmentation mechanisms.
Effects of dynamic chromatic and luminance contrast noise on the perception of coherent motion J. L. Barbur, J. A. Harlow and M. Fahle Applied Vision Research Centre, City University, Northampton Square, London, U.K. The purpose of this study was to investigate the independence of mechanisms involved in the detection of colour and luminance contrast defined motion in human vision. In addition, we also studied the effects of static or dynamic luminance contrast (LC) or chromatic contrast (CC) noise on thresholds for detection of either static or moving patterns defined by either colour or luminance contrast. Random dynamic noise was used to generate a texture of checks (check size 0.2o x 0.2o) by changing either the chromaticity or luminance of each texture element according to uniform luminance or chromaticity distributions. The textured pattern was generated in the centre of a uniform background field subtending 29ox 23o (Lb= 16 cd m-2, (x, y) - 0.305, 0.323). Dynamic noise was generated by varying randomly either the luminance or the chromaticity of each texture element, every 80 ms. The luminance range was specified as a % of background luminance. 1D chromatic contrast noise was specified by random sampling along a line of constant hue and 2D chromatic noise required random sampling over a specified region that varied in both hue and saturation. In addition to the ongoing, noise-generated, random motion of the texture elements, a subset of checks, defined either by a fixed luminance or chromatic contrast shifted systematically in the same direction so as to generate coherent movement. We measured the effect of LC and CC noise on thresholds for detection of either stationary or moving test patterns defined by either colour or luminance contrast. The results show that chromatic detection thresholds are largely unaffected by LC noise. On the other hand, the perception of coherent motion of colour defined, spatially random checks was completely disrupted by LC noise. Thresholds for detection of either stationary or moving luminance contrast defined patterns were affected equally by either the presence of 1D or 2D chromatic contrast noise. Once above detection threshold, the perception of luminance contrast defined coherent motion was found to be independent of chromatic contrast noise. The results suggest that unique mechanisms exist for the processing of luminance contrast defined motion, independently of any colour defined motion. The results also suggest that colour defined motion does not have its own first order motion detection mechanism and that the perception of colour defined motion is based largely on feature tracking.
Contrast sensitivity functions for temporal phase discrimination M.J. Wright Department of Human Sciences, Brunel University, Uxbridge, UB8 3PH, U.K. It was previously found that the perception of depth in simple shape-from-shading stimuli showed low-pass tuning to temporal frequency modulation, with a cutoff around 7Hz (Wright, M.J., ECVP 1998). It was proposed that the perception of luminance gradients depends on a low temporal frequency mechanism. This was tested using counterphase gratings (whose spatial gradient averaged over time is, of course, zero). The task was to discriminate between pairs of gratings that were temporally modulated either in phase (0 deg), or 90 deg apart in phase. That is, pairs of counterphase gratings, each in a circular window, were presented to the subject, and the task was to say whether they were modulating in phase or out of phase. At very low temporal frequencies this is trivially easy because one grating is at full contrast when the other is at zero contrast. However with increasing temporal frequency the discrimination becomes impossible long before the grating loses visibility. Contrast thresholds for "in-phase versus out-of-phase" discrimination of two counterphasing gratings were determined using a forced-choice method. The gratings were presented at a range of fixed contrasts spanning the threshold and 75% correct points were determined from fitted psychometric functions. Contrast sensitivity functions (CSF) for temporal phase discrimination were low- pass with a high frequency cutoff in the range 3Hz-7Hz depending on spatial frequency. The CSF for temporal phase discrimination was found to differ in shape from the temporal CSF for detection of a counterphase grating, as measured with the same displays. Analysis of these tuning functions supports evidence from a number of sources that temporal integration is required for the perception of the lightness or darkness of surfaces, and for the perception of luminance gradients.
Inter-ocular distance, enhanced disparity information and the control of prehension. Mark F. Bradshaw1, Paul B. Hibbard1, Simon J. Watt1, Rob van der Willigen1, Neil Stringer1, Ian R.L. Davies1, Nick Beagley2 and Andrew Willis2 1Department of Psychology, University of Surrey, Guildford, UK. Binocular disparity is considered pre-eminent in the control of reaching behaviour. One reason for this is that once suitably scaled (with an estimate of viewing distance and inter-ocular distance) it can specify metrical depth relationships within a scene. Such information is necessary in order to plan the transport and grasp phase of a reaching movement (i.e. absolute distance and size are required). If prehension is controlled via a binocular telepresence system, an opportunity arises to augment disparity information by increasing the inter-camera distance (ICD) beyond normal interocular distance, thereby increasing the magnitude (and range) of binocular disparity information. Whether an observer can take advantage of enhanced disparities to control reaching movements is unknown. Here we examine the effects of manipulating ICD on reaching movements with ICDs ranging from 6.5cm to 26cm. Typically sized, real world objects (both familiar and unfamiliar) were placed in a scene and reaching performance was assessed. An experimental sequence consisted of three blocks. The first and last block used a normal ICD/IOD of 6.5 cm whereas the middle block used an increased ICD. Larger than normal ICDs were found to disrupt reaching performance, with slower peak velocities and smaller grip apertures being observed. This was more pronounced for unfamiliar objects. Little evidence for learning was found.
Novel visual phenomena associated with zero persistence scanned displays. RJM Jones Centre for Human Sciences, DERA Farnborough, Farnborough, GU14 0LX. Although scanned displays are relatively common, all conventional display technologies have display persistence. Zero persistence scanned displays have recently become available, and early experiments using these displays have indicated several unexpected visual phenomena, for example, Kelly et al (Inv. Ophth. Vis. Sci. (Suppl) 39, S399, 1998) found that flicker sensitivity was reduced when using a zero persistence display, compared with a CRT display. We have performed experiments to investigate several aspects of vision using zero persistence displays, using a generic zero persistence display. Methods: Brightness matching was performed by presenting a scanned, uniform field on the display, except for a small central portion of the display, which was illuminated continuously. The brightness of this central portion was adjusted until the brightness matched that of the remaining part of the display. The luminance of the display ranged from 10 to 30 cd.m-2. Refresh rates were continuously variable up to 250Hz. Flicker sensitivity was measured using a similar approach to that of Corbett and White (Nature, 261, 689, 1976), except that the refresh frequency was also altered, and the display brightness remained constant (15 cd.m-2). Subjects were also asked to report any other unusual aspects of the display. Results: All subjects matched the display to a less bright continuous source. This effect was most noticeable at lower refresh rates (a factor of 1.2 at 60-70Hz), but was still present at over 100 Hz. The perception of flicker was found to be ‘slightly annoying’ at 37 Hz. Conclusions: Zero persistence displays seem to give a lower perceived sensitivity to brightness and flicker. Several mechanisms for these differences will be suggested.
Temporal delay affects the reach but not the grasp in natural prehension movements S.J. Watt and M.F. Bradshaw Department of Psychology, University of Surrey, Guildford, UK. Neuropsychological evidence suggests a dissociation between the systems involved in perceptual processes and those involved in the control of action (Milner and Goodale, 1995). We have reported previously that temporal delays between stimulus presentation and response produced inaccurate reaches to remembered distances whereas performance in a perceptual task was unaffected (Bradshaw et al., 1998). This is consistent with Milner and Goodale’s account as extrinsic object properties such as egocentric distance need to be updated rapidly as observers move around the world in order to accurately control the transport component of their reach. Here, we investigate the prediction that temporal delay similarly affects the grasp component even though intrinsic object properties, such as size and shape, do not change when observers move. 16 participants made open-loop reaches to three different sized objects placed at 25 and 40 cm along the midline, with a delay between target presentation (2 seconds) and response of 0, 1, 2 and 4 seconds. A MacReflex motion analysis system was used to record kinematic and spatial parameters of the movements. The results indicated that the reach component was increasingly affected as temporal delay was increased. Movements showed significantly slower peak wrist velocities, and detailed analysis of individual velocity profiles revealed that participants increasingly under-reached for the objects as delay increased. In contrast the grasp component was not affected by temporal delay. Maximum grip apertures remained consistent across all delay conditions and scaled only with object size. We conclude that extrinsic object information starts to deteriorate within 1 second, whereas intrinsic object information remains invariant over the range of delays tested. These results are inconsistent with the complete dissociation between the systems involved in perception and action, as proposed by Milner and Goodale.
Visual acuity with CRT and LCD displays PA Ward, SK King and JP Viveash DERA Centre for Human Sciences, Farnborough, Hants, UK. Due to their reduced weight, power requirements and size, liquid crystal displays (LCD) are becoming the technology of choice over cathode ray tube (CRT) displays in both the commercial and military markets. Several studies have compared visual performance on LCD and CRT systems, but always in a static environment. One key application of LCDs is in vehicles; an environment where vibration can be an important human factor. Previous work has shown that visual performance deteriorates when the observer is subject to vibration Viveash, J.P. et al (1994 Displays 15:74-75). This study examined visual performance using a simple acuity task performed on LCD and CRT displays whilst the head was vibrated. Six subjects carried out the experiment. During the acuity task the head was vibrated at 0, 14, 28 and 56Hz. The vibration was applied directly to the head through a chin rest fixed to a small electromagnetic vibrator. The subject viewed the displays from 6m. Both displays were matched for visual angle, contrast and luminance, and refresh rate (60Hz). The display consisted of a series of black illiterate E's on a white background (90cd.m-2). The illiterate E's ranged in visual acuities from 6/4 to 6/12 following normal Snellen rules. The results indicate a significant reduction in visual acuity using LCD displays for vibration frequencies approaching the refresh rate of the display system. Acuity was maintained on the CRT system even though image distortion was perceived. Thus under some conditions where the head is subject to vibration selection of display technology could affect visual performance. Acknowledgement British Crown Copyright 2000 / DERA
Speed discrimination for transparent motion J. M. Wallace & P. Mamassian Department of Psychology, University of Glasgow, 58 Hillhead Street, Glasgow G12 8QB. Motion transparency occurs when two motions in the same spatial location are perceived as distinct surfaces in motion, one sliding over the other. A number of studies have established the necessary conditions for the perception of motion transparency. In terms of Adelson & Movshon’s [1982. Nature, 300, 523-525] velocity space combination rule transparency occurs when combination fails. This can be due to a difference in spatial frequency, contrast, orientation and speed between the two components. Stoner, Albright & Ramachandran [1990. Nature, 344, 153-155] found that perception of motion transparency is dependent on the physics of transparency and by this account the motion system "knows" transparency. More recently Qian, Andersen & Adelson [1994. The Journal of Neuroscience, 14(12), 7357-7366] found that the phenomenon depends on ‘unbalanced’ motion signals that generate residual activation in MT/V5. To further explore the nature of transparent motion perception we measured speed discrimination thresholds in two tasks utilising standard signal-detection paradigms. In the motion coherency task, we presented two random dot kinematograms of translatory motions (left and right) in successive temporal intervals. Subjects were required to indicate which motion they perceived as being faster in a forced ‘left’ or ‘right’ response. In the motion transparency task, two translatory motions (left and right) were superimposed in a single temporal interval to simulate two transparent surfaces moving in opposite directions. Again, subjects were forced to respond ‘left’ or ‘right’ according to the motion they perceived as being faster. In both tasks different levels of noise, consisting of random dots moving at random speeds drawn from a fixed uniform distribution, were added to the stimuli while the speed ratio between standard and target stimuli was held constant. This was repeated for a range of speed ratios. We found that the relation between thresholds in the two tasks does not remain constant over the range of speed ratios. At large speed ratios, speed discrimination thresholds were higher in the transparency task than in the coherency task, while for low speed ratios thresholds were lower for transparent stimuli than the coherent stimuli. We discuss the implications of these results for models of motion transparency.
Towards a Real-time Implementation of a Biologically Motivated Model of Human Motion Detection J. Dale, A. Johnston, C. Benton, P. McOwan and M. Hodgetts Department of Psychology, University College London, WC2 6BT, Department of Computer Science, Queen Mary and Westfield College, London. E1 4NS, SIRA Technology Centre. BR7 5EH A complete model of visual motion processing requires a mathematical description which provides both a coherent computational strategy for motion extraction, and a computational role for components of the motion algorithm which can be identified with properties of neurones in visual cortex. The system as a whole should also reflect the performance of the human visual system in a variety of psychophysical motion tasks. We have been developing a model of visual motion processing which draws on known properties of neurones in the visual pathway, and which can be tested by deriving predictions for human performance for critical motion sequences and comparing the predictions against experimental data. The model represents the image luminance in a local region by means of a Taylor expansion. In order to compute velocity this representation is differentiated with respect to space and time. Speed and inverse speed estimates are calculated for a range of directions at the point of interest, allowing us to recover speed in terms of the amplitude of these radial functions and direction as a phase angle relative to some chosen reference frame. The mathematical model can also be applied to real video image sequences. Increases in the computational power available to computer vision systems is allowing us to develop a real-time implementation of the model, capable of extracting optical flow estimation from live video. Transferring this mathematical model of the biology into applicable technology provides further tests of the effectiveness of the approach, as well as a potentially useful computer vision system.
Pointing to stereoscopically defined planar surfaces. R.F. van der Willigen, M.F. Bradshaw and Paul B. Hibbard Department of Psychology, University of Surrey, Guildford, Surrey GU2 5XH, United Kingdom. Disparity-defined planar surfaces, oriented about a horizontal axis, are often perceived faster, and with more apparent slant, than those oriented about a vertical axis — the slant anisotropy (Rogers & Graham, 1983 Sci 1409-11; Gillam et al, 1984 Per & Psych 559-64). Does a similar anisotropy exist when we reach to stereoscopically defined surfaces? In two experiments we compared perceptual and visuo-motor indices of (i) slant magnitude estimation (ii) time-course for slant recovery. The stimuli were 10% random-dot surfaces with irregular boundaries that subtended 10 deg. They were presented on a single monitor viewed through a semi-silvered mirror set at 45 degrees to the median plane. The simulated viewing distance was 30 cm and the slants depicted ranged from 0 to ±60 degrees. Perceptual judgements were recorded by setting the orientation of a line to match the slant of the surface. Visuo-motor performance (pointing along the edge of the stimuli) was recorded by a MacReflex system. Pointing was ‘disparity open-loop’ in that only a monocular view of the hand was available during pointing. There were marked indivdual differences in terms of slant magnitude. No anisotropy was evident in either of the conditions for two of the eight subjects; four subjects showed a similar anisotropy in both conditions; and two observers showed a profound anisotropy in the perceptual condition which disappeared in the visuo-motor condition. In terms of time-course all subjects behaved similarly, perceptual settings built up over time reaching asymptote by six to eight secs. In the visuo-motor condition, however, slant was fully realised immediately and did not build-up with longer presentation durations. Slant reversals were more evident in the perceptual condition. For most subjects a similar response to surface slant is evident for both tasks. However, small but reliable differences between the tasks, were found for all subjects, in their temporal nature.
Colour discrimination performance at extreme gaze angles through visors coated with band-pass filters. Eric Liggins, Claire Roberts DERA Centre for Human Sciences, Farnborough, Hants, UK. Predicting colour discrimination performance at extreme gaze angles when viewing a scene through a protective band-pass filter coating on a large visor is non-trivial. Left and right eyes are provided with different spectral information about the scene, but judgements and ordering of hues can still be performed. The current experiment used the Farnsworth-Munsell 100 Hue test (FM-100) to assess binocular colour discrimination under a well-matched daylight illuminant (Macbeth Spectralight II) for both straight-ahead (0º) and extreme (30º) gaze angles. 10 subjects with normal colour discrimination participated in trials using a novel filtered laser protection visor consisting of dielectric interference coatings on a dyed polycarbonate substrate. Error scores were computed according to the method described by Farnsworth, and the 30º scores compared with the baseline (no filter) and the 0º conditions. Trials were also conducted using comparable neutral density filters to eliminate luminance-dependent effects. Results analysed in the usual manner showed that there was a highly significant (p<0.001) difference between the 0º and 30º angle viewing conditions through the filter, the colour discrimination at 30º being poorer. However, the error scores were typically skewed in each case (confirmed by calculation of skewness, g1) and the data were subsequently subjected to appropriate transformation (after KINNEAR, Vision Res. 10 423-433 (1970)) and re-analysis. The loss in colour discrimination does not appear to mimic any of the classical congenital or acquired colour vision defects, as shown by polar plots of FM-100 error scores. The implication for protective visors is that spectral transmission that changes with gaze angle must be taken into account if the visor is to be used when the wearer is performing visual tasks with colour-critical content. British Crown Copyright 2000 / DERA
Enhanced Disparity Information Under Telepresence Neil S. Stringer1, Mark F. Bradshaw1, Ian R.L. Davies1, Paul Hibbard1, Nick Beagley2 and Andrew Willis2 1Department of Psychology, University of Surrey, Guildford, GU2 5XH, UK. 2Centre for Human Sciences, Defence Evaluation Research Agency, Farnborough, Hants, GU14 0LX, UK. Stereopsis can provide a powerful cue to depth, however, binocular disparities vary with the inverse square of viewing distance (approximately) and fall below threshold within a few metres. Telepresence systems present the opportunity to augment disparities by increasing the inter-camera distance, effectively increasing the interocular distance of the user. Recovering a full metric representation of the field of view is computationally expensive, and it is likely that we have evolved to recover the lowest order geometrical representation sufficient to perform the task at hand. For many telepresence applications, the answer to the design question of what visual information to provide is influenced by the minimum bandwidth principle; provide just sufficient information for sufficient performance. With this in mind, performance on three tasks, differing in the minimum geometry required to perform them, were compared under normal and hyperstereopsis conditions. All tasks involved subjects positioning a "mobile" (with one degree of freedom) within a scene viewed, via remote cameras, on a monitor using a field sequential display and temporal shuttering glasses. Overall, it was concluded that subjects use different strategies for different perceptual tasks and that they can learn to use enhanced disparity information to increase the range over which these tasks are possible. |
||||||||
| webdesign by ablen | The AVA is a registered charity (No: 1049146) |