Perceptual saliency of points along the contour of everyday objects: a large-scale study

The aim of this large-scale study was to find out which points along the contour of a shape are most salient and why. Many subjects (N=161) were asked to mark salient points on contour stimuli, derived from a large set of line drawings of everyday objects (N=260). The database of more than 200,000 marked points was analyzed extensively to test the hypothesis, first formulated by Attneave (1954), that curvature extrema are most salient. This hypothesis was confirmed by the data: Highly salient points are usually very close to strong curvature extrema (positive maxima and negative minima). However, perceptual saliency of points along the contour is determined by more factors than just local absolute curvature. This was confirmed by an extensive correlational analysis of perceptual saliency in relation to ten different stimulus factors. A point is more salient when the two line segments connecting it with its two neighboring salient points make a sharp turning angle and when the 2-D part defined by the triplet of salient points is less compact and sticks out more.     

Identification of everyday objects on the basis of fragmented outline versions

Although Attneave (1954 Psychological Review 61 183 193) and Biederman (1987 Psychological Review 94 115-147) have argued that curved contour segments are most important in shape perception, Kennedy and Domander (1985 Perception 14 367-370) showed that fragmented object contours are better identifiable when straight segments are shown. We used the set of line drawings published by Snodgrass and Vanderwart (1980 Journal of Experimental Psychology: Human Learning and Memory 6 174-215), to make outline versions that could be used to investigate this issue with a larger and more heterogeneous stimulus set. Fragments were placed either around the 'salient' points or around the midpoints (points midway between two salient points), creating curved versus relatively straight fragments when the original outline was fragmented (experiment 1), or angular and straight fragments when straight-line versions were fragmented (experiment 2). We manipulated fragment length in each experiment except the last one, in which we presented only selected points (experiment 3). While fragmented versions were on average more identifiable when straight fragments were shown, certain objects were more identifiable when the curved segments or the angles were shown. A tentative explanation of these results is presented in terms of an advantage for straight segments during grouping processes for outlines with high part salience, and an advantage for curved segments during matching processes for outlines with low part salience.     

Contour-based object identification and segmentation: stimuli, norms and data, and software tools.

We summarize five studies of our large-scale research program, in which we examined aspects of contour-based object identification and segmentation, and we report on the stimuli we used, the norms and data we collected, and the software tools we developed. The stimuli were outlines derived from the standard set of line drawings of everyday objects by Snodgrass and Vanderwart (1980). We used contour curvature as a major variable in all the studies. The total number of 1,500 participants produced very solid, normative identification rates of silhouettes and contours, straight-line versions, and fragmented versions, and quite reliable benchmark data about saliency of points and object segmentation into parts. We also developed several software tools to generate stimuli and to analyze the data in nonstandard ways. Our stimuli, norms and data, and software tools have great potential for further exploration of factors influencing contour-based object identification, and are also useful for researchers in many different disciplines (including computer vision) on a wide variety of research topics (e.g., priming, agnosia, perceptual organization, and picture naming). The full set of norms, data, and stimuli may be downloaded from www.psychonomic.org/archive/.     

Contour-based object identification and segmentation: Stimuli,norms and data,and software tools

We summarize five studies of our large-scale research program, in which we examined aspects of contour-based object identification and segmentation, and we report on the stimuli we used, the norms and data we collected, and the software tools we developed. The stimuli were outlines derived from the standard set of line drawings of everyday objects by Snodgrass and Vanderwart (1980). We used contour curvature as a major variable in all the studies. The total number of 1,500 participants produced very solid, normative identification rates of silhouettes and contours, straight-line versions, and fragmented versions, and quite reliable benchmark data about saliency of points and object segmentation into parts. We also developed several software tools to generate stimuli and to analyze the data in nonstandard ways. Our stimuli, norms and data, and software tools have great potential for further exploration of factors influencing contour-based object identification, and are also useful for researchers in many different disciplines (including computer vision) on a wide variety of research topics (e.g., priming, agnosia, perceptual organization, and picture naming). The full set of norms, data, and stimuli may be downloaded fromwww.psychonomic.org/archive/.     

Identification of everyday objects on the basis of silhouette and outline versions

Line drawings of everyday objects were modified into silhouettes by filling-in the complete area enclosed by boundary contours, and outline versions were created by extracting the contours from the silhouettes. A large number of participants was asked to try to identify these silhouette and outline versions in experiment 1. Identifiability ranged from 0% to 100% correct responses with a large range in-between. Several kinds of errors and several reasons for difficulties with identification emerged in our data set. In experiment 2, we compared the original identification rates to those of inverted silhouettes (white figures on a black background), and in experiment 3 we compared the original identification rates of objects with filled-in holes or background parts to those of versions without filling-in. These stimuli and identification norms are useful for additional research on priming and context effects of object identification, neuropsychological deficits of object identification, and all kinds of studies with silhouettes where the role of top down knowledge could be of interest.     

Information concentration along the boundary contours of naturally shaped solid objects.

In this study of the informativeness of boundary contours for the perception of natural object shape, observers viewed shadows/silhouettes cast by natural solid objects and were required to adjust the positions of a set of 10 points so that the resulting dotted shape resembled the shape of the original silhouette as closely as possible. For each object, the observers were then asked to indicate the corresponding positions of the 10 points on the original boundary contour. The results showed that there was a close correspondence between the chosen positions of the points and the locations along the boundary contour that were local curvature maxima (convexities or concavities). This finding differs from that of Kennedy and Domander (1985 Perception 14 367-370), and shows that, at least for natural objects, the original hypothesis of Attneave (1954 Psychological Review 61 183-193) is valid--local curvature maxima are indeed important for the perception of shape.     

Segmentation of object outlines into parts: a large-scale integrative study

In this study, a large number of observers (N=201) were asked to segment a collection of outlines derived from line drawings of everyday objects (N=88). This data set was then used as a benchmark to evaluate current models of object segmentation. All of the previously proposed rules of segmentation were found supported in our results. For example, minima of curvature (i.e. locations along the contour where negative curvature takes an extreme value) were often used as segmentation points. The second point of a pair connected by a segmentation line often depended on more global shape characteristics such as proximity, collinearity, symmetry, and elongation. Based on these results, a framework is presented in which all of the previously proposed (and now empirically validated) segmentation rules or rules for part formation are integrated.     

Information along contours and object boundaries

F. Attneave (1954) famously suggested that information along visual contours is concentrated in regions of high magnitude of curvature, rather than being distributed uniformly along the contour. Here the authors give a formal derivation of this claim, yielding an exact expression for information, in C. Shannon's (1948) sense, as a function of contour curvature. Moreover, they extend Attneave's claim to incorporate the role of sign of curvature, not just magnitude of curvature. In particular, the authors show that for closed contours, such as object boundaries, segments of negative curvature (i.e., concave segments) literally carry greater information than do corresponding regions of positive curvature (i.e., convex segments). The psychological validity of this informational analysis is supported by a host of empirical findings demonstrating the asymmetric way in which the visual system treats regions of positive and negative curvature.     

Emotional valence modulates the preference for curved objects

Previous studies have shown that people prefer objects with curved contours over objects with sharp contours. However, those studies used stimuli that were mainly neutral in emotional valence. We tested here the interplay between visual features and general valence as positive or negative. After replicating curvature preferences for neutral objects, we used positive (cake, chocolate) and negative (snake, bomb) stimuli to examine if emotional valence-through response prioritisation-modulates the preference for curved objects. We found that people indeed preferred the curved versions of objects to the sharp versions of the same objects, but only if the objects were neutral or positive in emotional valence. There were no difference in liking for objects with negative emotional valence. This is evidence that the aesthetic response is adaptive, in this case prioritising valence over contour as demanded by the general semantic classification.     

Haptic identification of curved surfaces

In two experiments, the active haptic identification of three-dimensional mathematically welldefined objects is investigated. The objects, quadric surfaces, are defined in terms of the shape index, a quantity describing the shape, and curvedness, a quantity describing overall curvature. Both shape index and curvedness are found to have a significant influence on haptic shape identification . Concave surfaces lead to a larger spread in responses than convex ones. Hyperbolic surfaces show a slight tendency to be identified with more difficulty than elliptic ones. Surfaces with a high curvedness are identified more easily than those with a low curvedness. Results from experiments with constant and with random curvedness are indistinguishable . It is concluded that shape index and curvedness are psychophysically not confounded.     

Outline shape is a mediator of object recognition that is particularly important for living things

We assess the importance of outline shape in mediating the recognition of living and nonliving things. Natural objects were presented as shaded line drawings or silhouettes, and were living and nonliving things. For object decision (deciding whether an object may be encountered in real life) there were longer response times to nonliving than to living things. Importantly, this category difference was greater for silhouettes than for shaded line drawings. For naming, similar category and stimulus differences were evident, but were not as pronounced. We also examined effects of prior naming on subsequent object decision performance. Repetition priming was equivalent for nonliving and living things. However, prior presentation of silhouettes (but not shaded line drawings) reduced the longer RT to nonliving things relative to living things in silhouette object decision. We propose that outline contour benefits recognition of living things more than nonliving things: For nonliving things, there may be greater 2-D/3-D interpretational ambiguity, and/or they may possess fewer salient features.     

Illusory contours and spatial judgment

We investigated whether, in the human visual system, the mechanisms responsible for relative location judgments are the same when those judgments are made in the context of illusory contours and in the context of mentally joining two points. We asked subjects to align a dot with the oblique contour of an illusory surface or to align a dot with two markers at an oblique orientation. The systematic errors differed in direction for these two conditions. All the systematic errors were orientation dependent. The errors in aligning a dot with an illusory contour seem to be related to the asymmetrical shape of the single objects, which are able to induce an illusory contour, as well as figure-ground segregation.     

Rapid Integration of Contour Fragments: From Simple Filling-in to Parts-based Shape Description

Line drawings are easy to recognize, although the only information to the visual system is the contour itself. Starting from information theory and a theory of decomposition in parts, we investigated whether certain regions of such a contour are perceptually more relevant than others, using a deletion detection paradigm. In this paradigm, high detectability means poor contour integration, and vice versa. Regions of interest were curvature singularities, namely positive maxima (M+), negative minima (m?) and inflection points (I), of smooth, closed contours. In Experiment 1, we performed a first exploration of the detectability of deletions around these three types of curvature singularities. M+ deletions were easier to detect than the deletions around the other two singularities, a result that is explained using a post hoc combination of both mentioned theoretical frameworks. In Experiment 2, we replicated these findings using figure-background reversal, so thatthe same physical deletions could either be M+ or m?. Again, the M+ deletions were easier to detect than m? deletions. Although both types of singularities involve regions of high curvature changes, they differ in that mdeletions create gaps that concur with spontaneous segmentation.     

Surface geometry influences the shape of illusory contours

Geometric and neural models of illusory-contour (IC) synthesis currently use only local contour geometry to derive the shape of ICs. Work on the visual representation of shape, by contrast, points to the importance of both contour and surface geometry. We investigated the influence of surface-based geometric factors on IC shape. The local geometry of inducing-contour pairs was equated in stereoscopic IC displays, and the shape of the enclosed surface was varied by manipulating sign of curvature, cross-axial shape width, and medial-axis geometry. IC shapes were measured using a parametric shape-adjustment task (Experiment 1) and a dot-adjustment task (Experiment 2). Both methods revealed large influences of surface geometry. ICs enclosing locally concave regions were perceived to be systematically more angular than those enclosing locally convex regions. Importantly, the influence of sign of curvature was modulated significantly by shape width and medial-axis geometry: IC shape difference between convex and concave inducers was greater for narrow shapes than wider ones, and greater for shapes with straight axis and symmetric contours (diamond versus bowtie), than those with curved axis and parallel contours (bent tubes). Even at the level of illusory "contours," there is a contribution of region-based geometry which is sensitive to nonlocal shape properties involving medial geometry and part decomposition. Models of IC synthesis must incorporate the role of nonlocal region-based geometric factors in a way that parallels their role in organizing visual shape representation more generally.     

Three-dimensional features facilitate object recognition

Abstract

A priming paradigm was used to investigate the contribution of local features and global shape information in object recognition. Five types of incomplete forms were used as primes: (1) forms with both maxima (local curvature) and midsegments of edges present and aligned on the outline contour; (2) forms similar in global shape to the first version of stimuli, but with misaligned elements; (3) forms with only maxima; (4) forms with only midsegments of edges; and (5) forms containing 3D comer junctions (in Experiments 3 and 4). The target was an outline drawing of an object from which the incomplete prime was derived. Subjects were asked to name the target as rapidly as possible. Primes were presented at levels of contrast corresponding to identification thresholds, as well as above and below threshold levels (determined in Experiments 1 and 3). Facilitation effects relative to a neutral (no prime) condition occurred at threshold and above threshold for primes with aligned elements, forms with only maxima, and forms with only midsegments. Priming occurred only above threshold for forms with non-aligned elements. In Experiment 4 the presence of 3D local features increased the magnitude of priming relative to forms with midsegments and to forms with flat corners (in Experiment 2). This result suggests that 3D features facilitate object identification either because objects are stored in the form of volumetric entities or because 3D features are extracted early in visual processing.     

Color diagnosticity in object recognition

Does color influence object recognition? In the present study, the degree to which an object was associated with a specific color was referred to ascolor diagnosticity. Using a feature listing and typicality measure, objects were identified as either high in color diagnosticity or low in color diagnosticity. According to the color diagnosticity hypothesis, color should more strongly influence the recognition of high color diagnostic (HCD) objects (e.g., a banana) than the recognition of low color diagnostic (LCD) objects (e.g., a lamp). This prediction was supported by results from classification, naming, and verification experiments, in which subjects were faster to identify color versions of HCD objects than they were to identify achromatic versions and incongruent color versions. In contrast, subjects were no faster to identify color versions of LCD objects than they were to identify achromatic and incongruent color versions. Moreover, when shape information was degraded but color information preserved, subjects were less impaired in their recognition of degraded HCD objects than of degraded LCD objects, relative to their nondegraded versions. Collectively, these results suggest that color plays a role in the recognition of HCD objects.     

Color diagnosticity in object recognition.

Does color influence object recognition? In the present study, the degree to which an object was associated with a specific color was referred to as color diagnosticity. Using a feature listing and typicality measure, objects were identified as either high in color diagnosticity or low in color diagnosticity. According to the color diagnosticity hypothesis, color should more strongly influence the recognition of high color diagnostic (HCD) objects (e.g., a banana) than the recognition of low color diagnostic (LCD) objects (e.g., a lamp). This prediction was supported by results from classification, naming, and verification experiments, in which subjects were faster to identify color versions of HCD objects than they were to identify achromatic versions and incongruent color versions. In contrast, subjects were no faster to identify color versions of LCD objects than they were to identify achromatic and incongruent color versions. Moreover, when shape information was degraded but color information preserved, subjects were less impaired in their recognition of degraded HCD objects than of degraded LCD objects, relative to their nondegraded versions. Collectively, these results suggest that color plays a role in the recognition of HCD objects.     

Curvature and the visual perception of shape: theory on information along object boundaries and the minima rule revisited

Previous empirical studies have shown that information along visual contours is known to be concentrated in regions of high magnitude of curvature, and, for closed contours, segments of negative curvature (i.e., concave segments) carry greater perceptual relevance than corresponding regions of positive curvature (i.e., convex segments). Lately, Feldman and Singh (2005, Psychological Review, 112, 243-252) proposed a mathematical derivation to yield information content as a function of curvature along a contour. Here, we highlight several fundamental errors in their derivation and in its associated implementation, which are problematic in both mathematical and psychological senses. Instead, we propose an alternative mathematical formulation for information measure of contour curvature that addresses these issues. Additionally, unlike in previous work, we extend this approach to 3-dimensional (3D) shape by providing a formal measure of information content for surface curvature and outline a modified version of the minima rule relating to part segmentation using curvature in 3D shape.     

Visual judgment of similarity across shape transformations: evidence for a compositional model of articulated objects

A single biological object, such as a hand, can assume multiple, very different shapes, due to the articulation of its parts. Yet we are able to recognize all of these shapes as examples of the same object. How is this invariance to pose achieved? Here, we present evidence that the visual system maintains a model of object transformation that is based on rigid, convex parts articulating at extrema of negative curvature, i.e., part boundaries. We compared similarity judgments in a task in which subjects had to decide which of the two transformed versions of a 'base' shape-one a 'biologically valid' articulation and one a geometrically similar but 'biologically invalid' articulation-was more similar to the base shape. Two types of comparisons were made: in the figure/ground-reversal, the invalid articulation consisted of exactly the same contour transformation as the valid one with reversed figural polarity. In the axis-of-rotation reversal, the valid articulation consisted of a part rotated around its concave part boundaries, while the invalid articulation consisted of the same part rotated around the endpoints on the opposite side of the part. In two separate 2AFC similarity experiments-one in which the base and transformed shapes were presented simultaneously and one in which they were presented sequentially-subjects were more likely to match the base shape to a transform when it corresponded to a legitimate articulation. These results suggest that the visual system maintains expectations about the way objects will transform, based on their static geometry.     

The effects of presentation time on preference for curvature of real objects and meaningless novel patterns

Objects with curved contours are generally preferred to sharp‐angled ones. In this study, we aim to determine whether different presentation times influence this preference. We used images of real objects (experiment 1) and meaningless novel patterns (experiment 2). Participants had to select one of two images from a contour pair, curved and sharp‐angled versions of the same object/pattern. With real objects, the preference for curved versions was greatest when presented for 84 ms, and it faded when participants were given unlimited viewing time. Curved meaningless patterns were preferred when presented for 84 and 150 ms. However, in contrast to real objects, preference for meaningless patterns increased significantly in the unlimited viewing time condition. Participants discriminated poorly between the two versions (curved and sharp‐angled) of the meaningless patterns in the 84‐ and 150‐ms presentations (experiment 3). Therefore, in short times with meaningless patterns, participants selected mostly the curved version without being aware of the difference. In conclusion, presentation time, type of stimulus, and their interaction influence preference for curvature.