Underwater caustics disrupt prey detection by a reef fish.

Natural habitats contain dynamic elements, such as varying local illumination. Can such features mitigate the salience of organism movement? Dynamic illumination is particularly prevalent in coral reefs, where patterns known as 'water caustics' play chaotically in the shallows. In behavioural experiments with a wild-caught reef fish, the Picasso triggerfish (Rhinecanthus aculeatus), we demonstrate that the presence of dynamic water caustics negatively affects the detection of moving prey items, as measured by attack latency, relative to static water caustic controls. Manipulating two further features of water caustics (sharpness and scale) implies that the masking effect should be most effective in shallow water: scenes with fine scale and sharp water caustics induce the longest attack latencies. Due to the direct impact upon foraging efficiency, we expect the presence of dynamic water caustics to influence decisions about habitat choice and foraging by wild prey and predators.

Optimizing colour for camouflage and visibility using deep learning: the effects of the environment and the observer's visual system.

Avoiding detection can provide significant survival advantages for prey, predators, or the military; conversely, maximizing visibility would be useful for signalling. One simple determinant of detectability is an animal's colour relative to its environment. But identifying the optimal colour to minimize (or maximize) detectability in a given natural environment is complex, partly because of the nature of the perceptual space. Here for the first time, using image processing techniques to embed targets into realistic environments together with psychophysics to estimate detectability and deep neural networks to interpolate between sampled colours, we propose a method to identify the optimal colour that either minimizes or maximizes visibility. We apply our approach in two natural environments (temperate forest and semi-arid desert) and show how a comparatively small number of samples can be used to predict robustly the most and least effective colours for camouflage. To illustrate how our approach can be generalized to other non-human visual systems, we also identify the optimum colours for concealment and visibility when viewed by simulated red-green colour-blind dichromats, typical for non-human mammals. Contrasting the results from these visual systems sheds light on why some predators seem, at least to humans, to have colouring that would appear detrimental to ambush hunting. We found that for simulated dichromatic observers, colour strongly affected detection time for both environments. In contrast, trichromatic observers were more effective at breaking camouflage.

Feature matching and segmentation in motion perception.

We examined the role of feature matching in motion perception. The stimulus sequence was constructed from a vertical, 1 cycle deg-1 sinusoidal grating divided into horizontal strips of equal height, where alternate strips moved leftward and rightward. The initial relative phase of adjacent strips was either 0 degree (aligned) or 90 degrees (non-aligned) and the motion was sampled at 90 degrees phase steps. A blank interstimulus interval (ISI) of 0-117 ms was introduced between each 33 ms presentation of the stimulus frames. The observers had to identify the direction of motion of the central strip. Motion was perceived correctly at short ISIs, but at longer ISIs performance was much better for the non-aligned sequence than the aligned sequence. This difference in performance may reflect a role for feature correspondence and grouping of features in motion perception at longer ISIs. In the aligned sequence half the frames consisted of a single coherent vertical grating, while the interleaved frames contained short strips. We argue that to achieve feature matching over time, the long edge and bar features must be broken up perceptually (segmented) into shorter elements before these short segments can appear to move in opposite directions. This idea correctly predicted that overlaying narrow, stationary, black horizontal lines at the junctions of the grating strips would improve performance in the aligned condition. The results support the view that, in addition to motion energy, feature analysis and feature tracking play an important role in motion perception.

Stereoscopic and contrast-defined motion in human vision.

There is considerable evidence for the existence of a specialized mechanism in human vision for detecting moving contrast modulations and some evidence for a mechanism for detecting moving stereoscopic depth modulations. It is unclear whether a single second-order motion mechanism detects both types of stimulus or whether they are detected separately. We show that sensitivity to stereo-defined motion resembles that to contrast-defined motion in two important ways. First, when a missing-fundamental disparity waveform is moved in steps of 0.25 cycles, its perceived direction tends to reverse. This is a property of both luminance-defined and contrast-defined motion and is consistent with independent detection of motion at different spatial scales. Second, thresholds for detecting the direction of a smoothly drifting sinusoidal disparity modulation are much higher than those for detecting its orientation. This is a property of contrast-modulated gratings but not luminance-modulated gratings, for which the two thresholds are normally identical. The results suggest that stereo-defined and contrast-defined motion stimuli are detected either by a common mechanism or by separate mechanisms sharing a common principle of operation.

No local cancellation between directionally opposed first-order and second-order motion signals.

Despite strong converging evidence that there are separate mechanisms for the processing of first-order and second-order motion, the issue remains controversial. Qian, Andersen and Adelson (J. Neurosci., 14 (1994), 7357-7366) have shown that first-order motion signals cancel if locally balanced. Here we show that this is also the case for second-order motion signals, but not for a mixture of first-order and second-order motion even when the visibility of the two types of stimulus is equated. Our motion sequence consisted of a dynamic binary noise carrier divided into horizontal strips of equal height, each of which was spatially modulated in either contrast or luminance by a 1.0 c/deg sinusoid. The modulation moved leftward or rightward (3.75 Hz) in alternate strips. The single-interval task was to identify the direction of motion of the central strip. Three conditions were tested: all second-order strips, all first-order strips, and spatially alternated first-order and second-order strips. In the first condition, a threshold strip height for the second-order strips was obtained at a contrast modulation depth of 100%. In the second condition, this height was used for the first-order strips, and a threshold was obtained in terms of luminance contrast. These two previously-obtained threshold values were used to equate visibility of the first-order and second-order components in the third condition. Direction identification, instead of being at threshold, was near-perfect for all observers. We argue that the first two conditions demonstrate local cancellation of motion signals, whereas in the third condition this does not occur. We attribute this non-cancellation to separate processing of first-order and second-order motion inputs.

Spatial resolution and receptive field height of motion sensors in human vision.

We estimated the length of motion-detecting receptive fields in human vision by measuring direction discrimination for three novel stimuli. The motion sequences contained either (i) alternate frames of two differently oriented sinusoidal gratings; (ii) alternate frames of vertical grating and plaid stimuli or (iii) a vertical grating divided into horizontal strips of equal height, where alternate strips moved leftward and rightward. All three stimulus sequences had a similar appearance (of moving strips) and the task was to identify the direction of the central strip. For sequences (ii) and (iii), performance fell as the strip height decreased. Threshold height fell with increasing contrast up to about 20%, then levelled off at the critical strip height. Temporal frequency (1. 9-15 Hz) had no effect on the critical strip height. We argue that the receptive field length corresponds to twice this critical height. The length estimates were strikingly short, ranging from about 0.4 cycles at 3.0 cpd to 0.1 cycles at 0.1 cpd. These lengths agree well with the estimates derived at threshold by Anderson and Burr (1991, J. Opt. Soc. Am. A8, 1330-1339), and imply that the motion-sensing filters have very broad orientation tuning. These and other results are interpreted within the framework of a Gaussian derivative model for motion filtering. The sensitivity of motion filters to a broad range of orientations suggests a simpler view of how coherent plaid motion is processed.

Orientation sensitivity in human visual motion processing.

Orientation tuning of receptive fields is well documented in the spatial domain, but considerable variability exists amongst published estimates of orientation sensitivity of motion receptive fields. We used a two-frame motion sequence, in which one frame was binary noise and the other was a horizontally displaced and filtered version of the same noise field, to examine the orientation sensitivity of human motion mechanisms. Initially, orientations orthogonal to the direction of motion were removed from each filtered frame. Observers indicated perceived direction of motion in a single interval, binary choice task. D(max) was determined for different amounts of removed orientations, and found to remain constant across the removal of energy up to approximately +/-60 deg from vertical. In a second experiment, the orientations removed were now parallel to the direction of motion of the stimulus. D(max) fell as a cosine function with increasing removal of orientation information, in agreement with off-orientation looking or matched filtering predictions. The two experiments show the presence of mechanisms both broadly tuned and more narrowly tuned for orientation. A control experiment introduced an interstimulus interval between the two frames of our motion sequence. Performance on the direction discrimination task was severely degraded, indicating that the original results are not explicable in terms of a feature-tracking or long-range motion process. The presence of both broadly and narrowly tuned mechanisms implies multiple possible solutions to the processing of coherent plaid motion.

Does early non-linearity account for second-order motion?

A contrast-modulated (CM) pattern is formed when a modulating or envelope function imposes local contrast variations on a higher-frequency carrier. Motion may be seen when the envelope drifts across a stationary carrier and this has been attributed to a second-order pathway for motion. However, an early compressive response to luminance (e.g. in the photoreceptors) would introduce a distortion product at the modulating frequency. We used a nulling method to measure the distortion product, and then asked whether this early distortion could account for perception of second-order motion. The first stimulus sequence consisted of alternate frames of CM (100% modulation) and luminance-modulated (LM) patterns. Carriers were either 2-D binary noise (4 x 4 min arc dots) or a 4 c/deg grating, both modulated at 0.6 c/deg. The carrier was stationary while the phase of the modulating signal (LM alternating with CM) stepped successively through 90 degrees to the left or right. Motion was seen in a direction opposite to the phase stepping, consistent with early compressive distortion that induces an out-of-phase LM component into the CM stimulus. We measured distortion amplitude by adding LM to the CM frames to null the perceived motion. Distortion increased as the square of carrier contrast, as predicted by the compressive transducer. It also increased with modulation drift rate, implying that the transducer is time-dependent, not static. Thus early compressive non-linearity does induce first-order artefacts into second-order stimuli. Nevertheless this does not account for second-order motion, since perceived motion of second-order sequences (CM in every frame) could in general not be nulled by adding LM components. We conclude that two pathways for motion do exist.

Motion contrast: a new metric for direction discrimination.

The Adelson-Bergen energy model (Adelson, E. H., & Bergen, J. R. (1985). Spatiotemporal energy models for the perception of motion. Journal of the Optical Society of America A, 2, 284-299) is a standard framework for understanding first-order motion processing. The opponent energy for a given input is calculated by subtracting one directional energy measure (EL) from its opposite (ER), and its sign indicates the direction of motion of the input. Our observers viewed a dynamic sequence of gratings (1 c/deg) equivalent to the sum of two gratings moving in opposite directions with different contrasts. The ratio of contrasts was varied across trials. We found that opponent energy was a very poor predictor of direction discrimination performance. Heeger (1992). Normalization of cell responses in cat striate cortex. Visual Neuroscience, 9, 181-197) has suggested that divisive inhibition amongst striate cells requires a contrast gain control in the energy model. A new metric can be formulated in the spirit of Heeger's model by normalising the opponent energy (EL - ER) with flicker energy, the sum of the directional motion energies (EL + ER). This new measure, motion contrast (EL - ER)/(EL + ER), was found to be a good predictor of direction discrimination performance over a wide range of contrast levels, but opponent energy was not. Discrimination thresholds expressed as motion contrast were around 0.5 +/- 0.1 for the sampled drifting gratings used in our experiments. We show that the dependence on motion contrast, and the threshold of about 0.5, can be predicted by a modified opponent energy model based on current knowledge of the response functions and response variance of cortical cells.

Sub-pixel accuracy: psychophysical validation of an algorithm for fine positioning and movement of dots on visual displays.

Many visual experiments call for visual displays in which dots are plotted with very fine positional accuracy. Spatial hyperacuities and motion displacement thresholds can be as low as 5 sec arc. On computer graphics displays small angular displacements of a pixel can be obtained only with long viewing distances which impose a small field of view. To overcome this problem, we describe a method for positioning the centroid of a quadrel (a 2 x 2 block of pixels) with very high accuracy, equivalent to 0.4% of a pixel width. This enables dot displays to be plotted with high positional accuracy at short viewing distances with larger fields of view. We show psychophysically that hyperacuities can be measured with sub-pixel accuracy in quadrel displays. Motion displacement thresholds of 16 sec arc were measured in multiple-dot and single-dot displays even though the pixel spacing was 1.2 min arc. Quadrel displays may be especially useful in studies of optic flow and structure-from-motion which demand a fairly large field of view along with fine positional accuracy.

Global motion adaptation.

Image motion is initially detected locally. Local motion signals are then integrated across space in order to specify the global motion of objects or surfaces. It is well known that prolonged exposure to motion causes adaptation at the local motion level. We have investigated whether adaptation also occurs at the global motion level. We have devised a global motion stimulus (a random dot kinematogram) which has equal motion energy in opposite directions but nonetheless gives rise to global motion perception. At the local motion level, adaptation to this stimulus should cause equal adaptation in both directions and should not give rise to an aftereffect. Any aftereffect seen must therefore be attributable to adaptation at the global motion level. We find that following adaptation to this stimulus, judgements of the perceived direction of a test pattern are systematically biased towards the direction opposite to the adapting direction, suggesting that adaptation does occur at a level of visual processing at which global motion is represented.

Form overshadows 'opponent motion' information in processing of biological motion from point light walker stimuli.

The point light walker (PLW) has been taken to demonstrate the existence of mechanisms specialised in the processing of biological motion, but the roles of form and motion information in such processing remain unclear. While processing is robust to distortion and exclusion of the local motion signals of the individual elements of the PLW, the motion relationships between the elements - referred to as opponent motion - have been suggested to be crucial. By using Gabor patches oriented in relation to the opponent motion paths as the elements of the PLW, the influence of form and opponent motion information on biological motion processing can be compared. In both a detection in noise, and a novel form distortion task, performance was improved by orienting the elements orthogonally to the opponent motion paths - strengthening the opponent motion signal - compared to orienting them collinearly. However, similar benefits were found with static tasks presentations. Orienting the Gabor patches orthogonally to their opponent motion also benefits contour integration mechanisms by aligning neighbouring elements along the limbs of the PLW. During static presentations this enhanced form cue could account for all the changes in performance, and the lack of additional improvement in moving presentations suggests that the strengthened opponent motion signal may not be affecting performance. We suggest the results demonstrate the primacy of form information over that of opponent motion in the processing of biological motion from PLW stimuli.

First-order and second-order signals combine to improve perceptual accuracy.

The question of whether first-order (luminance-defined) and second-order (contrast-defined) stimuli can be combined in order to improve perceptual accuracy was examined in the context of two suprathreshold discrimination experiments, one spatial and the other temporal. The stimuli were either gratings of one type of image alone or else the sum of two gratings of the same orientation, spatial frequency, temporal frequency, and phase, but of different types. For both spatial frequency discrimination (static gratings) and speed discrimination (1-c/deg drifting gratings), performance was markedly better for a combined grating stimulus than predicted on the basis of independent processing of the two types of stimulus. But this was true only for stimuli of low contrast. Facilitation of discrimination performance occurred only in the contrast range where discrimination performance is contrast dependent. At higher contrasts, where performance has reached an asymptote for each type of pattern alone, there was no facilitation. The results suggest that first- and second-order stimuli, although believed by most researchers to be detected separately, can subsequently be combined in order to improve perceptual accuracy in conditions of low visibility.

What does the Ternus display tell us about motion processing in human vision?

The Ternus display is a moving visual stimulus which elicits two very different percepts, according to the length of the interstimulus interval (ISI) between each frame of the motion sequence. These two percepts, referred to as element motion and group motion, have previously been analysed in terms of the operation of a low-level, dedicated short-range motion process (in the case of element motion), and of a higher-level, attentional long-range motion process (in the case of group motion). We used a novel Ternus configuration to show that both element and group motion are, in fact, mediated solely by a process sensitive to changes in the spatial appearance of the Ternus elements. In light of this, it appears that Ternus displays tell us nothing about low-level motion processing, implying that previous studies using Ternus displays, for instance those dealing with dyslexia, require reinterpretation. Further manipulations of the Ternus display revealed that the orientation and spatial-frequency discrimination of the process underlying the analysis of Ternus displays is far worse than thresholds for spatial vision. We conclude that Ternus displays are analysed via a long-range motion, or feature-tracking, process, and that this process is distinct from spatial vision.