Ian van der Linde

DOVES: a database of visual eye movements.

DOVES, a database of visual eye movements, is a set of eye movements collected from 29 human observers as they viewed 101 natural calibrated images. Recorded using a high-precision dual-Purkinje eye tracker, the database consists of around 30 000 fixation points, and is believed to be the first large-scale database of eye movements to be made available to the vision research community. The database, along with MATLAB functions for its use, may be downloaded freely from http://live.ece.utexas.edu/research/doves, and used without restriction for educational and research purposes, providing that this paper is cited in any published work. This paper documents the acquisition procedure, summarises common eye movement statistics, and highlights numerous research topics for which DOVES may be used.

An efficient technique for revealing visual search strategies with classification images

We propose a novel variant of the classification image paradigm that allows us to rapidly reveal strategies used by observers in visual search tasks. We make use of eye tracking, 1/f noise, and a grid-like stimulus ensemble and also introduce a new classification taxonomy that distinguishes between foveal and peripheral processes. We tested our method for 3 human observers and two simple shapes used as search targets. The classification images obtained show the efficacy of the proposed method by revealing the features used by the observers in as few as 200 trials. Using two control experiments, we evaluated the use of naturalistic 1/f noise with classification images, in comparison with the more commonly used white noise, and compared the performance of our technique with that of an earlier approach without a stimulus grid.

A combinatorial study of pose effects in unfamiliar face recognition

The face inversion effect, evidence that humans possess a specialized system for face processing, and the (3/4) view advantage, evidence that a canonical viewpoint exists from which faces may be optimally recognized, are two commonly cited findings in the face processing literature. In this paper, the interaction of these effects is examined in a sequential matching paradigm in which unfamiliar faces are combinatorially randomized in pose across two dimensions (roll and yaw). Using large numbers of poses, trials and face stimuli, two experiments were conducted in which pose was either jointly or independently randomized between intervals. Results include that performance was modulated in a continuous fashion as each dimension was manipulated, that an offset-specific (3/4) advantage exists, that both specific study and test pose affect recognition, and that, for like offset, yaw rotation is more deleterious to performance than roll rotation. Response bias effects included that matched or reflective yaw led observers to employ a more liberal criterion.

Multiresolution image compression using image foveation and simulated depth of field for stereoscopic displays

Spatial contrast sensitivity varies considerably across the field of view, being highest at the fovea and dropping towards the periphery, in accordance with the changing density, type, and interconnection of retinal cells. This observation has enabled researchers to propose the use of multiple levels of detail for visual displays, attracting the name image foveation. These methods offer improved performance when transmitting images across low-bandwidth media by conveying only highly visually salient data in high resolution, or by conveying more visually salient data first and gradually augmenting with the periphery. For stereoscopic displays, the image foveation technique may be extended to exploit the additional acuity constraint of the human visual system caused by the focal system: limited depth of field. Images may be encoded at multiple resolutions laterally taking advantage of the space-variant nature of the retina (image foveation), and additionally contain blur simulating the limited depth of field phenomenon. Since optical blur has a smoothing effect, areas of the image inside the high-resolution fovea, but outside the depth of field may be compressed more effectively. The artificial simulation of depth of field is also believed to alleviate symptoms of virtual simulator sickness resulting from accommodation-convergence separation, and reduce diplopia.

How does aging affect the types of error made in a visual short-term memory 'object-recall' task?

This study examines how normal aging affects the occurrence of different types of incorrect responses in a visual short-term memory (VSTM) object-recall task. Seventeen young (Mean = 23.3 years, SD = 3.76), and 17 normally aging older (Mean = 66.5 years, SD = 6.30) adults participated. Memory stimuli comprised two or four real world objects (the memory load) presented sequentially, each for 650 ms, at random locations on a computer screen. After a 1000 ms retention interval, a test display was presented, comprising an empty box at one of the previously presented two or four memory stimulus locations. Participants were asked to report the name of the object presented at the cued location. Errors rates wherein participants reported the names of objects that had been presented in the memory display but not at the cued location (non-target errors) vs. objects that had not been presented at all in the memory display (non-memory errors) were compared. Significant effects of aging, memory load and target recency on error type and absolute error rates were found. Non-target error rate was higher than non-memory error rate in both age groups, indicating that VSTM may have been more often than not populated with partial traces of previously presented items. At high memory load, non-memory error rate was higher in young participants (compared to older participants) when the memory target had been presented at the earliest temporal position. However, non-target error rates exhibited a reversed trend, i.e., greater error rates were found in older participants when the memory target had been presented at the two most recent temporal positions. Data are interpreted in terms of proactive interference (earlier examined non-target items interfering with more recent items), false memories (non-memory items which have a categorical relationship to presented items, interfering with memory targets), slot and flexible resource models, and spatial coding deficits.

Visual Memory for Fixated Regions of Natural Images Dissociates Attraction and Recognition

Recognition memory for fixated regions from briefly viewed full-screen natural images is examined. Low-level image statistics reveal that observers fixated, on average (pooled across images and observers), image regions that possessed greater visual saliency than non-fixated regions, a finding that is robust across multiple fixation indices. Recognition-memory performance indicates that, of the fixation loci tested, observers were adept at recognising those with a particular profile of image statistics; visual saliency was found to be attenuated for unrecognised loci, despite that all regions were freely fixated. Furthermore, although elevated luminance was the local image statistic found to discriminate least between human and random image locations, it was the greatest predictor of recognition-memory performance, demonstrating a dissociation between image features that draw fixations and those that support visual memory. An analysis of corresponding eye movements indicates that image regions fixated via short-distance saccades enjoyed better recognition-memory performance, alluding to a focal rather than ambient mode of processing. Recognised image regions were more likely to have originated from areas evaluated (a posteriori) to have higher fixation density, a numerical metric of local interest. Surprisingly, memory for image regions fixated later in the viewing period exhibited no recency advantage, despite (typically) also being longer in duration, a finding for which a number of explanations are posited.

Object — Position Binding in Visual Short-Term Memory for Sequentially Presented Unfamiliar Stimuli

The effect of spatial position on visual short-term memory (VSTM) for sequentially presented objects has been investigated relatively little, despite the fact that vision in natural environments is characterised by frequent changes in object position and gaze location. We investigated the effect of reusing previously examined spatial positions on VSTM for object appearance. Observers performed a yes – no recognition task following a memory display comprising briefly presented 1/f noise discs (ie possessing spectral properties akin to natural images) shown sequentially at random coordinates. At test, single stimuli were presented either at original spatial positions, new positions, or at a fixed central position. Results, interpreted in terms of appearance and position preview effects, indicate that, where original spatial positions were reused at test, memory performance was elevated by more than 25%, despite that spatial position was task-irrelevant (in the sense that it could not be used to facilitate a correct response per se). This study generalises object – spatial-position binding theory to a sequential display scenario in which the influences of extrafoveal processing, spatial context cues, and long-term memory support were minimised, thereby eliminating the hypothesis that object priming is the principal cause of the ‘same-position advantage’ in VSTM.

Manual tapping enhances visual short‐term memory performance where visual and motor coordinates correspond

Visuo-manual interaction in visual short-term memory (VSTM) has been investigated little, despite its importance in everyday tasks requiring the coordination of visual perception and manual action. This study examines the influence of a manual action performed during stimulus learning on a subsequent VSTM test for object appearance. The memory display comprised a sequence of briefly presented 1/f noise discs (i.e., possessing spectral properties akin to natural images), wherein each new stimulus was presented at a unique screen location. Participants either did (or did not) perform a concurrent manual action (spatial tapping) task requiring that a hand-held stylus be moved to a position on a touch tablet that corresponded (or did not correspond) to the screen position of each new stimulus as it appeared. At test, a single stimulus was presented, either at one of the original screen positions, or at a new position. Two factors were examined: the execution (or otherwise) of spatial tapping at a corresponding or non-corresponding position, and the presentation of test stimuli either at their original spatial positions, or at new positions. We find that spatial tapping at corresponding positions elevates VSTM performance by more than 15%, but this occurs only when stimulus positions are matched from memory to test display. Our findings suggest that multimodal attentional focus during stimulus encoding (incorporating visual, spatial, and manual components) leads to stronger, more robust memory representations. We posit several possible explanations for this effect.

Foveated Analysis and Selection of Visual Fixations in Natural Scenes

The ability to automatically detect visually interesting regions in images has practical applications in the design of active machine vision systems. Analysis of the statistics of image features at observers gaze can provide insights into the mechanisms of fixation selection in humans. Using a novel foveated analysis framework, in which features were analyzed at the spatial resolution at which they were perceived, we studied the statistics of four low-level local image features: luminance, contrast, center-surround outputs of luminance and contrast, and discovered that the image patches around human fixations had, on average, higher values of each of these features than the image patches selected at random. Center-surround contrast showed the greatest difference between human and random fixations, followed by contrast, center-surround luminance, and luminance. Using these measurements, we present a new algorithm that selects image regions as likely candidates for fixation. These regions are shown to correlate well with fixations recorded from observers.

A study of human recognition rates for foveola-sized image patches selected from initial and final fixations on calibrated natural images

Recent years have seen a resurgent interest in eye movements during natural scene viewing. Aspects of eye movements that are driven by low-level image properties are of particular interest due to their applicability to biologically motivated artificial vision and surveillance systems. In this paper, we report an experiment in which we recorded observers’ eye movements while they viewed calibrated greyscale images of natural scenes. Immediately after viewing each image, observers were shown a test patch and asked to indicate if they thought it was part of the image they had just seen. The test patch was either randomly selected from a different image from the same database or, unbeknownst to the observer, selected from either the first or last location fixated on the image just viewed. We find that several low-level image properties differed significantly relative to the observers’ ability to successfully designate each patch. We also find that the differences between patch statistics for first and last fixations are small compared to the differences between hit and miss responses. The goal of the paper was to, in a non-cognitive natural setting, measure the image properties that facilitate visual memory, additionally observing the role that temporal location (first or last fixation) of the test patch played. We propose that a memorability map of a complex natural scene may be constructed to represent the low-level memorability of local regions in a similar fashion to the familiar saliency map, which records bottom-up fixation attractors.