Applied vision

Involving people in colouring their streets: From virtual images to a public art intervention

Grazia Caporusso, Stefano Feduzi, Laura Oppo, Maura Putzu and Mauro Murgia

Environmental changes would benefit of the involvement of people living in that environment. This would make them to feel an active part of the changing process. The present work is part of a participatory planning project conducted in a small Sardinian town, San Sperate. The aim of the research was to provide engineers and architects with population's judgments about a public art intervention aimed to colour the streets of their town. Pictures of the street with different asphalt colours were used as stimuli. Participants were 184 people resident in these streets, who were asked to rate virtual images representing 6 possible asphalt colours, on a likert scale from 0 to 10. The results evidenced that participants rated with higher scores those colours that fit well with the walls of the houses. Subjects'preferences were communicated to the engineers and architects, who planned the intervention on the basis of population's preferences. Concluding, we demonstrated that experts in visual perception, adopting a paradigm typical of the environmental psychology, can apply their knowledge in the social context.

Estimating weight and height based on mental norms: A cross-cultural study from Germany and Japan

Tobias M. Schneider, Claus-Christian Carbon and Heiko Hecht

The ability to exploit features of the human face to predict health and fitness can serve as an evolutionary advantage. Facial symmetry, averageness, and skin colour are known to influence health and fitness. However, numerous studies have found that human judgments are biased by diverse prototypes, cultural norms, and social heuristics. Are we able to use variables like weight and height to validly estimate human health and fitness by the mere inspection of the face, or could we expect that weight judgments on the basis of faces are biased by culturally shaped body prototypes? We used a cross-cultural approach with Caucasian versus Asian observers rating Caucasian and Asian faces. Participants used a heuristic based on their own cultural norm of weight (and height), as estimations were superior for faces of their own ethnicity. Accordingly, participants used a very similar heuristic for estimating weights of women versus men for both ethnicities based on the norm difference between the sexes of their own ethnicity. Taking into account that body estimations were historically limited to one's own ethnicity, such a simple heuristic makes sense, but leads to systematic biases in a globalized world.

Towards an emotional footprint: Non-verbal analysis of emotionally processed visual stimuli via posturography

Marius Raab, Nato Shengelia and Claus-Christian Carbon

The emotional impact of visual content is of great interest in fundamental research such as empirical aesthetics as well as in applied topics like the evaluation of usability and design appreciation. Measuring emotions elicited by visual stimuli, however, is a major challenge due to cognitive penetration in usual verbal reports, low temporal resolution in EDA, and general problems in revealing the valence of psychophysiological measures retrieved by EEG and fMRI. To overcome all these drawbacks, we propose posturography, which we employed with low cost commercial hard- and software. The utilized setting shows high temporal resolution (100Hz) and is capable of reliably tracking a person's balance control. A pilot (n=5) showed high and distinct postural responses to visual emotional stimuli. In a further study (n=21) we presented 23 distinct high-quality images (selected in a pre-study, rated for emotional intensity and value) in randomized order, each preceded and followed by an isoluminant mask, while the subject was standing on the balance board. The data reveal significant postural responses to emotional content with distinct patterns for several so-called base-emotions. Future research aims to identify the specific 'emotional footprints' for providing a reliable and valid instrument for measuring emotions elicited by visual stimuli.

Is there a common trait of aesthetic assessment or does it depend on the assessed content?

Uwe Christian Fischer, Claus-Christian Carbon, Johan Wagemans and M. Dorothee Augustin

People judge artworks with a view to different aesthetic dimensions [Augustin, Carbon, and Wagemans, 2011, Perception, 40]. But to what extent do they assess different art objects in a homogenous way, according to a general aesthetic trait? A latent multiple-state-model [Steyer, Schmitt, and Eid, 1999, European Journal of Personality, 13(5), 389-408] was used for different artworks, exploring possibilities to build a latent multiple-state-multiple-trait-model. Data of six different artworks were analyzed, which differed in their figural and abstract content (2 x landscape, 2 x persons, and 2 x abstract) and had been presented in random order to a mixed sample (N=304). To reduce complexity of the model we focused on three of five dimensions from an extended questionnaire [Augustin et al 2011]: Comprehensibility, Emotiveness and Pleasingness, with two indicator variables per dimension. The latent multiple-state-model with six pictures produced an acceptable fit, confirming the assessment structure. The intercorrelations of the equal latent dimensions were high for figural but not for abstract paintings. In consequence, it was not possible to realize an acceptable model with generalized single-traits for each state dimension within this study. Possible implications for the assessment of aesthetic objects are discussed.

A computational model for saliency detection based on probability distributions

Dominik Alexander Klein, German Martin Garcia and Simone Frintrop

The detection of salient items is a key property in human perception which is also of significant interest in technical fields such as computer vision. For example, the detection of salient items in web images facilitates tasks such as object segmentation and recognition (Liu et al, 2009, Transactions on Pattern Analysis and Machine Intelligence (PAMI), 33(2), 353-367). Here, we present a new way to compute saliency efficiently by using new methods for the representation of features and the computation of center-surround differences. Image regions are represented by multivariate normal distributions and compared with the Wasserstein metric. This is a well-known method to compare probability distributions and is especially suited to compute saliency since it considers also the similarity of feature values. We evaluated the method on psychophysical patterns as well as on a benchmark of natural images containing salient objects (Achanta et al, 2009, Proc. of Computer Vision and Pattern Recognition, 1597-1604) and show that the new approach outperforms nine state-of-the-art saliency detectors in terms of precision and recall.

Fast and accurate multi-scale keypoints based on end-stopped cells

Kasim Terzic, J.M.F. Rodrigues and J.M.H. Du Buf

Increasingly more applications in computer vision employ interest points. Algorithms like SIFT and SURF are all based on partial derivatives of images smoothed with Gaussian filter kernels. These algorithms are fast and therefore very popular. Our own multi-scale keypoint algorithm [Rodrigues and du Buf, 2006, BioSystems, 86, 75-90] is based on V1 end-stopped cells on top of complex cells with inhibition schemes to suppress responses along edges etc. Although producing good results, it is slow because of many filter orientations and scales. We therefore developed an improved algorithm which is much faster, because instead of using big filter kernels (Gabor filters as simple cells) at coarse scales, we apply a Gaussian pyramid and do all filtering in the frequency domain. Stability and localisation are improved by automatic scale selection and by smoothing response maps prior to detecting local maxima. Extensive benchmarking concerning repeatability, precision and computing speed showed that the improved algorithm compares to or even outperforms most algorithms from computer vision. Since the code will be made publicly available, our new keypoint algorithm can be applied in advanced biological vision models and in computer vision applications. [Projects: PEst-OE/EEI/LA0009/2011, NeFP7-ICT-2009-6 PN: 270247]

Relationship between imitating an expert's motion and improved motor skills

Yuya Akasaka and Miyuki G. Kamachi

Imitation is an important function in human development, and imitating an expert's motion is considered an efficient way for beginners to improve their motor skills. In this study, we investigated the relationship between an improvement in beginners'skill of throwing darts and how closely they imitated the expert's motion they observed. Fourteen male participants, all right-handed, participated in the experiment, and all were considered as beginners since they had little experience with darts. The experiment was conducted over two days. On the first day, all participants threw darts in their own way. On the second day, 7 of them were instructed to imitate an expert's motion by watching a video, and threw darts. The other 7 participants threw darts without observing the expert's motion. The motion similarity of the expert and both beginner groups was analyzed using an Angular Metrics for Shape Similarity (AMSS). The result showed that the score of the observing group was higher than that of the other group. In addition, the closer the beginner came to imitating the expert's motion, the more skilled the beginner became at throwing darts. We suggest that the degree of motion similarity has a bearing on the improvement of motor skills.

Glare and Neural Contrast: Scene content controls the neural contrast response

Alessandro Rizzi and John McCann

We collected magnitude estimates of scene lightness in white, black and average (half-white/half-black) surrounds [Rizzi et. al., JSID, 2007] and then calculated their retinal luminance using Vos and van den Berg's glare spread function [CIE,1999]. Each of the three sets of lightness magnitude estimates fit straight-line log luminance functions with different slopes. Gray appearances in a white surround fit the cube root of scene luminance, and the log of retinal luminance. Scattered light from intra-ocular glare accounts for the change in shape of the psychometric function. In a white surround, a retinal luminance range of 30:1 generates the entire appearance range from white to black. In a black surround, that white/black appearance range requires 100,000: 1 retinal range, while the average surround requires a100:1 retinal range. Scene content, namely, the amount of white area, controls the appearance of grays. The three scene-dependent psychometric functions of retinal luminance shows that the neural contrast processing of retinal luminance responds to the spatial content of the scene. The greater the retinal glare, the greater the neural contrast processing producing higher apparent contrast.

Motion processing deficits in migraine, contrast sensitivity and what is relevant noise for a visual motion task

Alexandra Shepherd, Heidi Beaumont and Trevor Hine

Background: There are conflicting reports concerning the ability of people with migraine to detect and discriminate visual motion. Enhanced coherent motion detection, yet impaired global motion discrimination, has been reported and the latter attributed to functional differences in extrastriate cortex in migraine. Previous studies have used different types of motion displays and none has adequately assessed other parameters that could affect performance. Methods: Motion-direction detection, discrimination and relative motion thresholds with random dot displays were compared from participants with migraine with and without aura and from control participants. Potentially relevant visual covariates were included (contrast sensitivity: CS; acuity; stereopsis; visual discomfort; visual stress; visual triggers; dyslexia). Results: For each motion task, both migraine groups were less accurate than the control group. Both migraine groups also had impaired CS, greater visual discomfort, visual stress and visual triggers. Only CS correlated with performance on each motion task, it also mediated performance. Dynamic visual noise (twinkling dots at 20 HZ) added to the relative motion task improved performance for the migraine group only, compared to static visual noise. Conclusions: Impaired performance on certain motion tasks can be attributed to impaired CS early in the visual system. There were additional differences between the groups for global and relative motion thresholds embedded in noise, suggesting changes in extrastriate cortex. Tasks that can distinguish the effects of noise on performance at different levels of the visual system are recommended in future basic and clinical research.

High-precision psychophysics on the iPad

Michael Dorr, Luis Lesmes, Long To, Zhong-Lin Lu and Peter Bex

Precise assessment of visual function, e.g. contrast sensitivity, typically requires large and expensive laboratory setups and long testing times. Digital display devices such as TFTs often suffer from limited luminance and spatio-temporal resolution, and CRTs with video attenuators are therefore still being used for precise stimulus control. In order to overcome these limitations, we implemented the quick CSF method (Lesmes et al., 2010, Journal of Vision 10(3):17, 1-21) on an Apple iPad2. We developed a novel and efficient algorithm to extend the display's contrast resolution from a minimum luminance increment of 1.2% at mean luminance to <0.2%. The integrated camera was utilized to monitor viewing distance, eliminating the need for a chin rest. Using repeated measurements from four observers with normal/best-corrected and optically blurred (+4D) vision, we showed that reliable CSF estimates with peak sensitivities of 0.6-0.8% can be obtained in 5-7 minutes and that these estimates are indistinguishable from those obtained with a CRT and video attenuator setup. With our system, it becomes feasible to rapidly and precisely measure contrast sensitivity inside and outside the laboratory, for example to collect population data, improve eye care delivery in medically underserved areas, and closely monitor the progression of visual neuropathologies.

A simple clinical test for assessing performance in areas V1 and MT

Linda Bowns

The problem: to produce a simple clinical test that measures normal performance in areas V1 and MT. Method: The test uses optimised moving plaid stimuli that requires an observer to encode the orientation and motion of each component (V1), and then recombine the components to form pattern motion (MT). The test controls for all known variables that effect motion direction and importantly cannot be done using second order motion. The test uses a method of constant stimuli and a forced choice task. Observers decide if the moving plaid is moving clockwise or anticlockwise relative to a fixed line. The test results reported here cover a four octave range of spatial frequencies at different directions at 45 deg. intervals around a circle. Results: the observers in experiment 1 report that the task is simple and perform close to 100% correct. Experiment 2 showed that the results remain similar under informal and non-calibrated viewing conditions. Conclusions: the test would provide a simple screening test invaluable to many researchers and clinicians where cortical visual problems have been implicated, e.g. dyslexia, Alzheimer's disease, dementia, autism; or for screening purposes where visual information plays a very important role, e.g. drivers, pilots, or air traffic controllers.

Eye-movement planning during flight maneuvers

Lewis Chuang, Frank Nieuwenhuizen and Heinrich H. Buelthoff

How are eye-movements planned to access relevant visual information during flight control? From the cockpit perspective, there are two classes of visual information that are relevant for flight control. First, the changing visuals of the external world provide direct perceptual feedback on how the pilot's command of the control stick is affecting the aircraft's current position, orientation and velocity. Second, flight instruments provide abstracted and specific values - on factors such as the aircraft's compass bearing and vertical speed - that have to be continuously monitored, in order for the global objective of certain maneuvers (e.g., turns) to be achieved. Trained pilots have to coordinate their eye-movements across this structured visual workspace (i.e., outside view and instruments) to access timely and task-relevant information. The current work focuses on providing descriptions of these planned eye-movements. Eye-movements were recorded of pilots in a high-fidelity flight simulator (100° field-of-view) whilst they performed specific flight maneuvers. Fixation durations and transitions between the individual instruments and aspects of the external environment are represented as network graphs. This allowed us to formally describe the sources of information that were relied on across the different tasks and to compare actual performance to expert predictions.

Blavigator: a navigation aid for blind persons

Joao Jose, Manuel Moreno, Javier Pinilla-Dutoit, J.M.F Rodrigues and J.M.H. Du Buf

Blavigator (blind navigator) is a vision aid for blind and visually impaired persons. It supports local navigation by detecting walkable paths in the immediate vicinity of the user. It guides the user for centering on the path. It also detects obstacles, both static and moving, in front of the user and just beyond the reach of the white cane, such that the user can be alerted. The user can choose between modulated sounds and synthesised speech for path centering and obstacle alerts. Local navigation works both indoor (corridors) and outdoor (sidewalks etc). Global navigation, for wayfinding in neighbourhoods and in buildings, is also possible, provided that a detailed geographic information system is available. Different technologies are used for localising the user: outdoor GPS reception, indoor triangulation of WiFi access points, and visual recognition of landmarks. Already working in realtime on a netbook computer with a simple webcam, the system is cheap, simple to install and maintain, and reliable with a user-friendly interface. The system is being tested in collaboration with ACAPO, the Portuguese association of blind and amblyopes. The ultimate goal is to use the system on a mobile phone with a built-in camera. [Projects: PEst-OE/EEI/LA0009/2011, NeFP7-ICT-2009-6 PN: 270247, RIPD/ADA/109690/2009; PhD grant SFRH/BD/82310/2011]

CORF: A computational model of a simple cell with application to contour detection

George Azzopardi and Nicolai Petkov

We introduce a computational model of a simple cell, which combines the responses of model LGN cells with center-surround receptive fields (RF). We call it Combination of RFs (CORF) model. We use simulated reverse correlation to demonstrate that the RF map of the CORF model can be divided into elongated inhibitory and excitatory regions, typical of simple cells. Besides orientation selectivity, the CORF model exhibits contrast invariant orientation tuning, cross orientation suppression and response saturation, which are observed in simple cells. These three properties are, however, not possessed by the Gabor function (GF) model, which has gained particular popularity as a computational model of a simple cell. We use two public data sets of images of natural scenes with associated ground truth to compare the CORF and the GF models in a contour detection task, which is assumed to be the primary biological role of simple cells. In this task, the CORF model outperforms the GF model (RuG dataset: t(39) = 4.39, p < 10^{-4}, Berkeley dataset: t(299) = 3.88, p < 10^{-4}). The proposed CORF model is more realistic than the GF model as it shares more properties with simple cells and it is more effective in contour detection.

Cortical Masking and Action Learning

Martin Thirkettle, Peter Redgrave, Kevin Gurney, Tom Walton, Ashvin Shah and Tom Stafford

Understanding the relationship between between our actions and sensory events in the environment is a fundamental skill. Dopaminergic (DA) neurons in the basal ganglia are thought to be the neural substrate of the action-outcome learning mechanism. Responsible for detecting relationships between motor output and sensory input, allowing behavioural contingencies to be detected and new actions to be discovered. We have previously shown that visual input which cannot access the DA neurons directly via subcortical pathways and must instead first submit to early cortical processing are less effective for such action learning (Thirkettle, et al, 2011, Perception, 40 ECVP Supplement, 33). It remains to be seen, however, if subcortical visual processing alone is sufficient to support action-outcome learning. Therefore we tested this with a motor learning task using reinforcing signals masked from cortical processing using constant-flash-suppression. Results show that masked reinforcing signals, which are unavailable to conscious access, are capable of supporting action-outcome pairing, but are less effective than unmasked stimuli. These results confirm our previous finding that subcortical visual pathways provide the bulk of the sensory input for action-outcome learning mechanisms, however they also suggest that the results of cortical processing contribute strongly to this system.

The spot that decides on the Gestalt: Gestalt psychology in the context of blindspot perception

Theresia K. Reiter and Claus-Christian Carbon

Gestalt psychology provides a prominent fundament for understanding higher-level perception. Several rules, the so called Gestalt principles that describe how multiple visual features are perceptually integrated to a whole, enable mainly post hoc explanations clarifying interpretations of specific configurations as 'Gestalts'. The good Gestalt principle refers to the simplest interpretation building the Gestalt, but as simplicity is not perfectly defined the principle's predictive quality for ambiguous configurations is low. To specify corresponding processes, we presented 63 ambiguous configurations to our participants projecting the visual information essential for solving the ambiguity monocularly on the retina's blindspot. The employed configurations comprised three sets built up of images requiring filling-in mechanisms based on geometric (GP), experience-driven (EP) and culture-driven (CP) principles, respectively. Participants described how they perceived the image, de facto how they filled-in the non-existing visual information. GP was mainly used as validation condition to check the individual's typical capability of filling-in, CP as test for culture-driven and EP for experience-driven filling-in mechanisms, respectively. Participants showed clear general capabilities of filling-in (GP=54.8%) with high test-retest reliabilities (Rs=.87). In CP and EP filling-in often yielded specific solutions for highly familiar images indicating factors of personal and cultural experience for solving ambiguity

An investigation into size-speed illusions in railway level crossing collisions

Helen Clark, John Perrone, Robert Isler and Samuel Charlton

The prevalence of collisions between motor vehicles and trains at railway level crossing junctions has been a high-profile issue for many years in New Zealand and other countries. Errors made in judging a train's speed could possibly be attributed to motorists being unknowingly subjected to a size-speed illusion. Leibowitz (1985, American Scientists, 95, 558-562) maintained that a large object seems to be moving slower than a small object travelling at the same speed. Support has been provided for Leibowitz's theory from studies using simple shapes on a screen. However the reasons behind the size-speed illusion remain unknown and there is no experimental evidence that it applies in the approaching train situation. To investigate these issues we tested observers'relative speed estimation performance for a train and a car approaching at a range of speeds and distances, in a simulated environment. A psychophysical method (2AFC) was used to establish the PSE for the two different-sized vehicles. The data show that participants significantly underestimate the speed of the train, compared to the car. A size-speed illusion seems to be operating in the case of the approaching train in our simulation and may also be a factor in some railway level crossing collisions.

Perception of synthetically generated interactive human emotional body expressions

Nick Taubert, Andrea Christensen, Dominik Endres and Martin A. Giese

The generation of reactive highly realistic human movements in computer graphics is a difficult problem. We investigate the perception of naturalism and emotional style from computer-generated interactions between human avatars, generated with a novel algorithm based on a hierarchal Gaussian process latent variable model [Taubert et al., 2011, KI-2011, 330-334], whose latent space includes dimensions for emotional style variation (neutral, happy, angry, fearful, sad). The model produces in real-time reactive human movement, such as handshakes, allowing for the separate control of emotional style (e.g. 'fearful'vs. 'angry'), the type of motion and individually specific motion style. In addition, the model is capable of morphing between motion styles in a framework that generates reactive motion online. In order to validate our approach, we conducted a 'Turing test': a psychophysical study where human observers rate generated and natural emotional handshakes. We find that the distinction between natural and synthesized movements is difficult for human observers, supporting the adequacy of our method for the synthesis of stylized interactive human movements for applications in computer graphics and robotics. Acknowledgements: EC-FP7-ICT-projects TANGO-249858, and AMARSi-248311, DFG project GI 305/4-1, and the Hermann Lilly Schilling Stiftung.

Matching unfamiliar faces under ecological conditions: Factors of stimulus similarity and feature exchange

Benedikt Emanuel Wirth and Claus-Christian Carbon

Face matching is considered a highly sophisticated process. However, unfamiliar face matching seems to be substantially based on iconic information and thus susceptible to pictorial changes [Carbon, 2008, Perception, 37, 801-806]. Amongst others, effects of inversion and negativation, important variables for basic research have mainly been investigated. Meanwhile variables most relevant for applied research have not been addressed as frequently. We investigated the accuracy of unfamiliar face matching in the applied context of passport controls under controlled conditions. Participants were asked to identify depicted persons by the photos of their passports. Three kinds of features were manipulated systematically: Paraphernalia, internal features, and external features. Averaged across all conditions, participants showed an accuracy of 82.6%, which does not meet expectations regarding the reliability of passport controls. Effects of changed features on accuracy, sensitivity, and reaction times were measured. Results showed no impairing effect of changing internal features, a small but significant effect of changing paraphernalia, and a large effect of changing external features. The data pattern provides relevant information for understanding specific processes of unfamiliar faces as well as for revealing typical problems in applied contexts.

Spotting expertise in the eyes: billiards knowledge revealed by gaze shifts and detected through machine learning

Sofia Crespi, Carlo Robino, Giuseppe Boccignone, Mario Ferraro and Claudio De' Sperati

In sports, as in other activities and knowledge domains, expertise is highly valuable. We assessed whether expertise in billiards could be deduced from the pattern of eye movements when experts, compared to novices, are simply observing a filmed billiard match with no specific task, and ad-hoc occluded billiard shots with a trajectory prediction task. Experts performed better than novices, both in terms of accuracy and response time. In addition experts did not need to mentally extrapolate the hidden part of the ball trajectory, a strategy used by novices, but fixated selectively the diagnostic points of the ball trajectory, in accordance with the metrical system used by billiard professionals to calculate the shot coordinates. Since eye movements of experts contain the signature of billiard expertise, we applied machine learning to classify expertise from the oculomotor behavior. By using saccade amplitude and direction as features, a Relevance Vector Machine classified correctly which group - novice or expert - the observers belonged to, with an accuracy of 83% and 77% respectively for the match and the shots. Our data suggest that the signature of expertise in billiards is hidden in the ocular scanpath, and that it can be accurately detected at the individual level.

How elite batters coordinate the movement of their eyes and head to anticipate interception

David Mann, Wayne Spratford and Bruce Abernethy

The performance of elite athletes in fast ball-sports approaches the limits of visual-motor behaviour, yet very few studies have sought to uncover how these athletes use visual gaze to guide and control motor action. Visual gaze strategies change commensurate with the level of interceptive skill [Land & McLeod, 2000, Nature Neuroscience, 3(12), 1340-1345], though the current understanding of motor behaviour is predicated on the strategies of lesser-skilled performers. In this study we recorded the eye and head movements of two of the world's best cricket batters and show that they use very specific gaze strategies which separate them from lesser-skilled batters, and from all batters examined previously. In particular, the elite batters use central vision to watch the bat hit the ball, a task beyond the capabilities of lesser-skilled batters. They do so (i) by producing large predictive saccades to anticipate the locations of ball-bounce, and bat-ball contact, and (ii) by precisely coupling the movement of their head to the flight-path of the ball, ensuring that the ball remains in a consistent egocentric direction relative to their head. Eye and head movements are a hallmark of expertise in batting; the findings are used to inform coaching practice in fast ball-sports.

Reversal rate of ambiguous fidure after glare conditioning stimulus

Angel Kurtev

The reversibility of an ambiguous figure seems to be determined by both bottom-up, stimulus-related manipulation, and top-down, attentional/cognitive, visual processing. Glare inducing stimuli modify the retinal input while causing fluctuation of attention at the same time, and therefore can affect the frequency of reversal. In order to study this possibility we measured the reversal rate of a Necker cube while intermittently presenting glare-inducing stimuli - bright homogenous field or series of flashes. The experiments were performed under mesopic viewing conditions with the Necker cube outlined in a positive contrast on a dark background. The experimental procedure was controlled by SuperLab Stimulus Presentation System (STP100W) connected to stimulator module STM 100C and a stroboscope TSD122A (Biopac, Inc.). The results obtained, while confirming the high interpersonal variability and the presence of fast and slow observers, showed that glare increases the reversal rate of the perceived orientation. Most of this outcome may be ascribed to nonspecific arousal changes. Considering the responses of myopic and emmetropic subjects revealed a trend for more pronounced change in the rate of reversal in emmetropes as compared to myopes. This difference may be a result of the higher sensitivity to glare in the myopic subjects under mesopic conditions.

Understanding bird egg patterning: assessing visual camouflage using visual difference and edge-detection models

Karen A. Spencer, Keri Langridge, Graeme D. Ruxton and P. George Lovell

Bird eggs can show a high degree of patterning, which is thought to enhance crypsis and thus avoid predation. Most studies to date have used manipulations of egg patterning to assess predation as a proxy for camouflage. Here we present the results of a study where Japanese quail were given a choice of four laying substrates (white, yellow, red or black sand). Laying locations were recorded for a period of 5 days. Egg detectability was quantified by measuring the amount of the egg to substrate boundary identified by the Canny edge detection algorithm. We further assessed laying choices by the estimating visual difference between the egg and each of the available substrates (To et. al., 2010). The chosen laying position minimised the amount of edge contour found by the edge-detector. Somewhat counter-intuitively, we found that the chosen substrate also tended to be the least visually similar to the egg. We also detected a significant negative correlation between the chosen substrate albedo and the degree of maculation (darker speckling) of eggs. These results suggest that egg camouflage may operate via disruptive colouration and that females may use the degree of maculation as a cue to optimising individual laying positions.

Action videogame playing can improve visual-motor control without affecting vision

Li Li, Rongrong Chen and Jing Chen

We examined how action videogame playing affects visual-motor control using a manual control task in which participants used a joystick to keep a blob centered on a large display as its horizontal position was randomly perturbed. Six naive Non-Videogame Players were trained with an action videogame (Mario Kart Wii, Nintendo), and six were trained with a strategy videogame (Roller Coaster Tycoon III, Atari) for 1-2 hours a day for 10 hours in total. Their performance on the manual control task was measured before the training, after 5-hour training, and after 10-hour training, and their contrast sensitivity function (CSF) was measured before and after training. For the group trained with the action videogame, the RMS error of their performance on the manual control task decreased by 14% (SD: 8%) after 5-hour training and by 20% (SD: 6%) after 10-hour training, and their overall control response (gain) increased by 24% (SD: 11%) after 5-hour training and by 32% (SD: 15%) after 10-hour training. The improvement sustained when they were retested on the manual control task between 2-4 months later. In contrast, no change of RMS or gain was observed for the group trained with the strategy videogame. For both groups, no change in CSF was found. We conclude that action videogame playing can improve visual-motor control without affecting vision.

Aligning fundus images via the automatic extraction of retinal surface vessels

Sven Holm and Niall McLoughlin

Extraction of retinal surface vessels allows for fast registration of human fundus images. Two unsupervised, automatic blood vessel extraction algorithms were implemented. The first approach enhanced the surface vessels by applying Gabor filters [Oloumi et al, 2007, Conf Proc IEEE Eng Med Biol Soc., 6451-6454]. The second approach combines multiple preprocessing steps to reduce the influence of noise within the images (such as the central light reflex) [Marín et al, 2011, IEEE Trans Med Imaging, 30, 146-158]. Both approaches resulted in similar accuracy (~94%) when applied to the DRIVE database of fundus images [Staal et al, 2004, IEEE Trans Med Imaging, 23, 501-509]. Spectroscopic fundus images were then aligned using the outputs of each approach. In particular, the automatically generated vessel masks for each fundus image were shifted both vertically and horizontally relative to a reference vessel mask. For each possible shift, the absolute difference between the two masks was computed. The translation that generates the minimal difference between the two masks was used to align the raw images. Thus, translations, resulting from minor eye tremors and movements, were corrected. Spectroscopic fundus images can then be used to calculate the relative oxygenation of different retinal compartments.