Skeletal representations of shape in human vision: Evidence for a pruned medial axis model

Abstract

A representation of shape that is low dimensional and stable across minor disruptions is critical for object recognition. Computer vision research suggests that such a representation can be supported by the medial axis—a computational model for extracting a shape's internal skeleton. However, few studies have shown evidence of medial axis processing in humans, and even fewer have examined how the medial axis is extracted in the presence of disruptive contours. Here, we tested whether human skeletal representations of shape reflect the medial axis transform (MAT), a computation sensitive to all available contours, or a pruned medial axis, which ignores contours that may be considered “noise.” Across three experiments, participants (N = 2062) were shown complete, perturbed, or illusory two-dimensional shapes on a tablet computer and were asked to tap the shapes anywhere once. When directly compared with another viable model of shape perception (based on principal axes), participants' collective responses were better fit by the medial axis, and a direct test of boundary avoidance suggested that this result was not likely because of a task-specific cognitive strategy (Experiment 1). Moreover, participants' responses reflected a pruned computation in shapes with small or large internal or external perturbations (Experiment 2) and under conditions of illusory contours (Experiment 3). These findings extend previous work by suggesting that humans extract a relatively stable medial axis of shapes. A relatively stable skeletal representation, reflected by a pruned model, may be well equipped to support real-world shape perception and object recognition.

A dual role for shape skeletons in human vision: perceptual organization and object recognition

“…Indeed, incorporating a skeletal model into off-the-shelf convolutional neural networks (CNNs) significantly improves their performance on visual perception tasks (Rezanejad et al, 2019). Similarly, behavioral research with humans has shown that participants extract the skeleton of 2D shapes (Firestone & Scholl, 2015;Kovács, Fehér, & Julesz, 1998;Psotka, 1978), even in the presence of border perturbations and illusory contours (Ayzenberg, Chen, Yousif, & Lourenco, 2019). Other research has shown that skeletal models are predictive of human object recognition (Destler, Singh, & Feldman, 2019;Lowet, Firestone, & Scholl, 2018;Wilder, Feldman, & Singh, 2011), even when controlling for other models of vision .…”

Skeletal descriptions of shape provide unique perceptual information for object recognition

“…Indeed, shape skeletons may offer a concrete formalization for the oft poorly defined concept of global shape. Skeletal descriptions exhibit many properties ascribed to global shape percepts such as relative invariance to local contour variations 22,35 . Moreover, there exist many methods by which to compare skeletal structures, such as by their hierarchical organization 61 or using distance metrics (as used here) 27 , thereby allowing for a quantitative description of shape similarity.…”

“…Accumulating evidence suggests that the skeletal structure of objects is extracted by the primate visual system during shape perception. In particular, behavioral studies have shown that human participants extract the skeletons of different 2D shapes 30–34 and those skeletal structures remain relatively stable across border disruptions resulting from perturbations or illusory contours 35 . Increasingly, studies have shown that skeletal structures may be represented in three dimensions (3D) within an object-centered reference frame.…”

Perception of an object’s global shape is best described by a model of skeletal structure in human infants

“…Evidence from human adults demonstrates that participants spontaneously extract the skeletons of 2D shapes, even in the presence of incomplete contours (Ayzenberg, Chen, Yousif, & Lourenco, 2019;Firestone & Scholl, 2014). Moreover, a 3D skeletal model predicted participants' object similarity and category judgments, even when controlling for other models of vision .…”