AVA & Colour Group Post-graduate Meeting

AVA-CG Post-graduate meeting 2nd November University of Newcastle.

Organised by:
Marina Bloj (m.bloj@brad.ac.uk) or Andrew Welchman (a.e.welchman@ncl.ac.uk).

Applied Vision Association (http://www.dmu.ac.uk/ava/) Colour Group (http://www.colour.org.uk)

AVA / Colour Group Postgraduate Meeting

2nd November 2001, University of Newcastle

There was an invited talk given by Professor Roger Watt

The prize for best poster was awarded to: Vandana Mehta
The prize for best paper was awarded to: Jennifer Skillen  

PAPERS

Characteristics of the accommodation system in myopia and emmetropia.
Fuensanta Vera-Diaz,Dpt. Of Optometry, University of Bradford  

Despite extensive myopia research, fundamental questions underlying the mechanism of myopisation remain unresolved. Many studies have tried to identify the factors which induce the increase in axial length associated with myopia development. One area of interest is in the relationship between myopia and the accommodation response. Previous results have shown that inaccuracies of the accommodation response occur in some myopes during and after prolonged near work. The aim of the study described is to examine the characteristics of myopic individuals that exhibit inaccurate accommodation responses. The results suggest that inaccuracies in the accommodation response occur predominantly during the progression of myopia. The implications of the findings in relation to models of myopia development are discussed.

A comparison of real life and computer generated motion in depth
V.L. Tuck, J.M. Harris. Department of Psychology, University of Newcastle upon Tyne, UK.  

When an object moves directly towards or away from us in depth the motion signals generated in each eye are equal but opposite. Researchers often use computer generated stimuli to simulate real world 3D motion on a 2D screen by presenting equal and opposite motion signals to each eye. Whilst this type of presentation simulates motion in depth, it is not identical to object motion in the real world. So, are observer responses to real life 3D motion identical to their responses to computer simulations of 3D motion? We have chosen to explore this question by looking at direction discrimination responses to monocular and binocular presentations of motion in depth. To compare observer responses over the two situations we conducted a series of experiments using real world motion in depth and computer simulations of the same 3D motion stimuli. In the real world experiments, an LED was moved directly towards and away from observers. Small movements were made at points in front of, around and behind a stationary fixation LED and observers were asked to report the direction of movement. Computer replications of the experiment were then carried out using the same observers. Comparisons in performance between viewing with one eye and two were made for both sets of stimuli. So far, two main findings have emerged. 1) With real life stimuli, binocular viewing is superior to monocular viewing for a range of positions in front of and around fixation whilst that advantage is lost for stimuli presented behind fixation. This is not the case for computer simulations as binocular performance deteriorates as the motion gets further from fixation and closer to the observer. 2) Thresholds for direction discrimination are lower for real life motion in depth than for computer presented motion in depth. This holds for binocular and monocular viewing. To help explain these findings we have identified several cues present in the real life stimuli but not in the computer simulations. A further series of experiments, to ascertain how useful these cues are in the direction discrimination of 3D motion, is now underway.

Quality of life, visual function and ocular health in South Asian and Caucasian adults in Bradford, UK
R Patel(1), S Pardhan (2), J Gilchrist (1) A Hobbiss (3)
1. Department of Optometry, University of Bradford, BD7 1DP UK
2. Department of Optometry and Ophthalmic Dispensing, Anglia Polytechnic University, CB1 1PT
3. Clinical Epidemiology Unit, University of Bradford, BD7 1DP UK  

Purpose: South Asians constitute a majority of the ethnic minorities in the UK. In Bradford, they comprise one-sixth of the total inner city population. This study aimed to investigate the quality of life, ocular disease, visual and functional vision of older Asian and Caucasians visiting an optometric practice in Bradford.
Method: All patients over the age of 45 years who visited the practice over a 12-month period were invited to participate. A comprehensive assessment of the visual function using a battery of tests including log MAR distance acuity, near acuity, face recognition and contrast sensitivity was carried out on the first eligible 100 South Asians and 100 Caucasians. Measurements were obtained for each eye as well as binocularly in the absence and presence of a glare source. Stereopsis and central visual field using an Amsler chart were also measured. A diagnosis of the ocular health was made. Data on the quality of life (QOL) were obtained by patients completing the NEI-VFQ 25 questionnaire.
Results: Univariate and multivariate Regression analysis were employed to determine tests that were significantly associated with QOL scores. Non-parametric tests (Mann-Whitney) tested for differences in the questionnaire score by race (p<0.05). Analysis of the data from the two racial groups showed that, although Asians were significantly younger and had significantly better visual function (functional vision and contrast sensitivity with and without glare), their quality of life scores were worse compared to those of Caucasians who showed worse visual function. In fact Asians showed significantly lower scores for items relating to driving, role dependency and ocular pain. Asians had significantly higher frequency of cataracts and lower frequencies of age-related macular degeneration. A large number of Asians (39%) were unaware of the presence of their cataract.
Discussion: Issues other than visual function are responsible for the relatively lower quality of life scores in the Asian group compared to Caucasians. These issues and their clinical implications are discussed.  

Keynote Lecture
Visual factors as possible cause of the Paddington Rail Crash of 5/10/99
Profesor Roger Watt Department of Psychology, University of Stirling

The Importance of Spatial Scale in Determining Illusions of Orientation and Curvature
J. Skillen¹, D. Whitaker¹, Ariella Popple² and P.V. McGraw¹.
¹Department of Optometry, University of Bradford. UK; ² Gatsby CNU, UCL, UK.

Interactions between 1st-order (luminance-defined) and 2nd-order (contrast-defined) information have been widely demonstrated in the orientation domain. The Fraser twisted cord illusion is an example where the global, 2nd-order structure is pulled towards the orientation of the local, 1st-order structure. When 1st-order orientation exceeds approximately 10 degrees a reversal of the Fraser effect to a Zollner effect occurs, where the 2nd order structure is pushed away from the orientation of the 1st order structure. We investigated interactions between 1st and 2nd-order information using a Fraser twisted cord illusion within the domains of orientation and curvature. Observers were presented with horizontal 1st-order gratings whose contrast was modulated by a Gaussian envelope elongated along the length of the grating. The orientation and curvature of the 1st-order carrier was systematically varied and its effect on the perceived orientation and curvature of the 2nd-order envelope was established. Spatial frequency effects were examined by varying the frequency of the carrier within an envelope of fixed size. We show that the relative spatial scale of carrier and envelope represents a decisive factor in determining the magnitude and direction of interaction effects and conclude that the perceived 2nd-order structure of a stimulus is biased by the properties of the 1st-order structure in a manner that depends on relative, rather than absolute spatial scale.

The Effect of Reduced Colour Depth on the Colour Reproduction of Web Images
Efthimia Bilissi, Imaging Technology Research Group,University of Westminster, Watford Road, Harrow, Middlesex HA1 3TP

Accurate colour reproduction of images viewed across the Internet can be a difficult task due to problems, such as variable viewing conditions. The standard RGB colour space (sRGB) has been proposed to overcome this problem without the need for special software applications. Different graphics adapter bit-depth settings also affect the number of displayable colours on a web browser. For this reason, the 216 colours ‘web safe’ palette has been introduced. To understand the effect of device characteristics on the process of image reproduction via the web, the characterisation of some common devices was carried out. This created a controlled environment whereby experiments could be conducted with reproducible results. Device characterisation also enabled the accurate conversion of scanned images to sRGB by using a step- by-step process that followed the specification of the colour space. The investigation of colour reproduction on a Cathode Ray Tube display under different bit-depth settings has been conducted using a suitable test target, which was converted to the sRGB colour space and also to the ‘web safe’ (216 colours) palette. These two images were displayed under 8-bit and 24-bit depth of the graphics card adaptor. Colorimetric measurements concerning colour differences were performed and evaluated. Experimental investigation of the true colour (24-bit colour depth) image while displayed under 8-bit setting via different browsers followed. The outcome of the investigations was evaluated and combined to determine the relationship between the different parameters described above and their effect on image quality.

Non-classical mediators of chromatic contrast induction
Kit Wolf and Anya Hurlbert, University of Newcastle

The chromatic contrast between a figure and its surround is a major determinant of its colour appearance. The magnitude of the induced colour change depends non-linearly on the properties of the immediate surround, and is also influenced by the chromaticities of remote surfaces up to 10 degrees distant from the figure (Wachtler & Sejnowski, 2001). We investigated a possible role of such remote fields in ‘stabilising’ the colour of the figure against local changes in chromatic context. When the colour of the background alone is varied in the direction of increasing L-cone excitation, the colour change induced in the center-square is less than when both background and remote surfaces are L+ shifted. By varying the S-cone contrast between the remote squares we tested the prediction that increasing the number of distinct surfaces in the scene should improve colour constancy (suppress the local contrast induction). For most observers there was no difference between conditions with two or eight distinct colours in the remote context. We also show that segmenting the remote surfaces from the figure by increasing their luminance reduces the contrast suppression. A final experiment attempted to localise the neural mechanisms involved. A Wheatestone viewer was used to present the stimuli haploscopically, the figure and remote context being presented either to the same eye or to different eyes. The effect of the remote context was identical in both cases, pointing to a binocular site.

Reference:
Wachtler, T., Albright TD, Sejnowski TJ. (2001) Nonlocal interactions in color perception: Nonlinear processing of chromatic signals from remote inducers. Vis Res. 41, 1535-1546

POSTERS

Early neural activity and its role in the development of structural and functional organisation of retinal ganglion cells in the turtle
Vandana Mehta and Evelyne Sernagor Department of Neuroscience, School of Neuroscience and Psychiatry, University of Newcastle upon Tyne.

Mature retinal ganglion cells (RGCs) with distinct functional properties have distinct morphologies, leading to the postulate that such structure-function correlation manifests in developing RGCs. Spontaneously arising retinal neural activity is an epigenetic factor critical for the maturation of RGC receptive fields (Sernagor and Grzywacz, 1995). Based on these findings it is hypothesised that embryonic spontaneous activity influences the outgrowth of RGC dendritic trees. To test this hypothesis, the structure-function correlates of developing turtle RGCs were investigated. The study was performed from Stage 23 (S23) when light responses emerge, until 15 weeks post-hatching. RGCs were labelled with Lucifer Yellow using patch electrodes while their responses to light were recorded. The results reveal that immature receptive fields expressed at S23-25 are not homogeneous and do not show a direct correlation with the layout of their dendritic trees. Changes in dendritic morphology during development were also investigated in RGCs back-labelled with the tracer horseradish peroxidase. Following intense dendritic proliferation up to S25, dendritic arbours undergo remodelling, and pruning at ulterior stages. Pruning is significant in large field RGCs. It is concluded that the lack of homogeneity expressed by immature receptive fields originates from sparse synaptic inputs onto the RGCs rather than from the spatial layout of their growing dendritic trees. When spontaneous activity is chronically blocked in vivo with curare, a cholinergic nicotinic antagonist, RGC dendritic growth is inhibited. Conversely, persistent and enhanced spontaneous activity caused by rearing hatchling turtles in the dark (Sernagor and Grzywacz, 1996) induces abnormally strong dendritic proliferation in large field RGCs. It is concluded that immature spontaneous activity promotes dendritic growth in developing RGCs.

Fourier statistics of fruit in foliage - a match with human contrast sensitivity?
C. A. Párraga, T. Troscianko, D. J. Tolhurst
Department of Experimental Psychology, University of Bristol,

The human visual system encodes luminance as a Fourier band-pass function and chrominance as a low-pass function. It is tempting to think that this imbalance reflects a property of the natural world. However, it has been found that there is no such difference between the Fourier luminance and chrominance spectra of a wide variety of natural scenes. If a main task of the chromatic channels is to encode information about fruit in foliage, we asked whether the Fourier properties of such scenes are different from the set of images analysed earlier. We therefore obtained 91 digital images of fruit, and other scenes, using a custom-calibrated digital camera. We analysed these scenes to obtain their luminance and red-green chrominance spectra. The results indicate that scenes containing fruit and foliage have power spectra which are more in keeping with human contrast sensitivity. Chrominance images of these scenes have steeper spectral slopes than luminance images. This is particularly the case when the red fruit occupies more than 10% of the total area of the image, corresponding to relatively close-up viewing. These results suggest that the spatiochromatic properties of human vision optimise the neural representation of scenes containing fruit and leaves.

The Effect of Speed and Time on Induced Motion
K.J. German, J.M. Harris. Department of Psychology, University of Newcastle upon Tyne, Newcastle upon Tyne, UK.

Purpose: In a previous experiment, induced motion in depth was directly compared with induced lateral motion at various speeds. The results indicated no significant effect of type of motion (lateral vs motion in depth) but there was a significant effect for speed: the induction effect decreased with increasing speed. However, time covaried with speed. Here we explore whether this fall off in magnitude is due to time, speed, or both factors. Two experiments are reported: (1) time remains constant whilst manipulating speed; (2) speed remains constant whilst manipulating time. The hypotheses are: (a) if time is the determining factor, then experiment 1 should show no significant decline in magnitude when speed is increased, whereas experiment 2 should show a change in magnitude as time is altered; (b) if speed is the determining factor, then experiment 1 should show a change in magnitude as speed is increased, whereas experiment 2 should show no significant change in magnitude.
Methods: Observers viewed 3 vertically aligned dots oscillating in one of 2 sets of directions: (a) continuously side to side (lateral motion), or (b) continuously forward and back (motion in depth). Observers fixated the central target dot and performed a nulling task to make the target dot appear stationary. The amount of motion required to null any movement in the target was taken as a measure of the induced motion perceived.
Results: For experiment1, perceived induced motion ranged from 0% to 90%. The effect of speed was significant: with time constant, the magnitude of induced motion decreased as a function of increasing speed. The effect of type of motion and the interaction of speed and type of motion were not significant. For experiment 2, perceived induced motion ranged from 50% to 80%. The effect of time was not found to be significant: with speed constant, the magnitude of induced motion did not alter significantly as time increased. The effect of motion and the interaction of time and motion were also not significant.
Conclusion: For the parameters used, the results suggest that speed is the predominant factor in the perception of induced motion, whether the motion is lateral or in depth. The results are discussed in relation to the theories of induced motion.

Higher-order statistical structure in colour-opponent representations of natural images
R. J. Paltridge & M. G. A. Thomson
Colour & Imaging Institute, University Of Derby, Kingsway House East, Kingsway, Derby, UK.

It has been suggested that natural scenes must surely display very significant colorimetric redundancy even when down-projected onto computational approximations of the three colour-opponent channels of human vision. For example, the power spectra of red/green-channel natural-image data are hardly more low-pass than that of the corresponding luminance-channel data, a result which seems at odds with human colour contrast-sensitivity (this is essentially low-pass) and which is unlikely to be affected significantly by optical/neural degradation at the retina [1]. Is this high-spatial-frequency data ignored because of limited neural capacities, or is it really predictable from data in the other colour channels? The present work tries to answer this question by analyzing the statistics of a newly developed, high quality database of hyperspectral images collected in a collaborative project between Aston, Derby and Keele universities. A variety of opponent-channel representations are tested (including spaces derived through characteristic vector analysis), and techniques borrowed from higher-order statistics are used to assess the cross-channel dependencies in a manner which does not depend on the exact power spectra in the three channels. The results imply that data in the three channels become increasingly correlated with increasing spatial frequency. This finding is in line with results reported by Ruderman et al.[3], who observed that whilst spatial edges tend to be co-located in all three channels, the directions of the colour changes at those edges are often independent. We discuss (a) the possible physical basis of these findings in terms of simple generative scene models, and (b) the consequences of these findings for the ‘scaleinvariance’ property that is often assumed to hold for natural scenes.

[1] C. A. Parraga, G. Brelstaff, T. Troscianko and I. R. Moorhead. Color and luminance information in natural scenes, J. Opt.Soc.Am.A, 15(3):563-569, March 1998.
[2] T. A. Yates, M. G. A. Thomson,R. J. Paltridge, S. Westland and S. M. Wuerger. Higher-order Statistics and the Coloured Structure of Natural Scenes, ECVP2001, 2001.
[3] D.L. Ruderman, T.W. Cronin and C.C. Chiao. Statistics of Cone Responses to Natural Images : Implications for Visual Coding, J.Opt.Soc.Am.A, 15(8):2036-2045, August 1998.