The perception of shape and curvedness from binocular stereopsis and structure from motion

Abstract

The integration of binocular stereopsis and kinetic depth was measured for two distinct aspects of 3-Dstructure: (1) shape index, which is a measure of scale-independent structure, and (2) curvedness, which is a measure of scale-dependent structure. We found that motion contributes significantly more to judged shape index than it does to judged curvedness, and stereo contributes significantly more to judged curvedness than it does to judged shape index. This suggests that the differences in the relative contribution of motion and stereo reported here occurred because these two sources do not equally specify the scale-dependent and scale-independent aspects of surface structure. Furthermore, these results seem to be inconsistent with integration models in which the different visual cues all initially contribute to the same single representation of 3-Dstructure.Our perception ofan object or surface consists ofmany different properties, such as its distance, size, and shape. Although the first two of these attributes can be defined in a relatively straightforward manner, there is not uniform agreement about what it means to perceive the shape of an object or surface (e.g., it could be its depth, orientation, curvature, or some combination of properties). Distinguishing between these possibilities is particularly important when considering the different types of retinal information available to specify three-dimensional (3-D) shape. For example, theoretical analyses demonstrate that while relative depth can be recovered directly from binocular disparities, it is difficult, if not impossible, to recover this information directly from shading or texture. In the case oftexture and shading, it is more likely that the visual system instead recovers relative slant or curvature without first computing local depth. Because of these potential differences, a full understanding of3-D shape perception must entail not only a description of those properties represented by the visual system but also the degree to which each is specified by the various optical sources of information. The purpose ofthe experiments presented here was to examine the integration of different surface properties from two of the more salient sources of 3-D structure: motion and binocular stereopsis.To provide an example of some of the candidates for local shape descriptors, consider a quadric surface patch of the form (1)

The stereoscopic anisotropy affects manual pointing

“…However, they are not inconsisten t with the idea that the tasks accessed separate representations using, for example, separate representations of slant and of absolute depth (c.f. Tittle and Perotti, 1997), or independent representations supporting perception and the control of action (Milner and Goodale, 1995). However, if multiple representations are involved, then they are clearly equally affected by the stereoscopic anisotropy.…”

“…Only when determining the absolute size and shape of an object is a full speci cation of metric Euclidean structure required. Tittle and Perotti (1997) similarly proposed that different tasks make use of different depth representations but, rather than the hierarchical framework discussed by Glennerster et al (1996), suggest that multiple independent representations of depth structure are produced. Therefore, in their model a representation which describes the three-dimensiona l positions of points may exist side-by-side with a representation of higher order properties such as surface slant, or curvature.…”

“…Watanabe et al, 1999;Bradshaw et al, 2000;Bradshaw and Hibbard, 2001). The representations proposed by Tittle and Perotti (1997) provide information such as depth curvature and shape index which is suitable for perceptual tasks like the recognition of three-dimensiona l shape (de Vries et al, 1993). However, not all tasks require rich high-level surface descriptions of this type.…”

Trajectory forms as a source of information about events

“…A quantity that has been studied in the structure-frommotion literature is curvature (Koenderink, 1990;Perotti et al, 1998;Tittle & Perotti, 1997). There, it has been used to describe the structure of perceived surfaces.…”

Brain Areas Involved in Perception of Biological Motion

“…Humans are adept at detecting weak coherent motion amidst a background of incoherent motion (Raymond, 1994;van de Grind, Koenderink, van Doorn, Milders, & Voerman, 1993;Williams & Sekuler, 1984; but see Barlow & Tripathy, 1997), are very accurate at judging the direction in which objects are moving (Gros, Blake, & Hiris, 1998;De Bruyn & Orban, 1988;Pasternak & Merigan, 1984;Ball & Sekuler, 1982), and, under optimal conditions, are keenly sensitive to slight differences in the speed at which objects are moving (Chen, Bedell, & Frishman, 1998;McKee & Welch, 1989;De Bruyn & Orban, 1988;Orban, De Wolf, & Maes, 1984). In addition, motion provides a potent source of information for specifying the 3-D shapes of objects (Tittle & Perotti, 1998;Sperling, Landy, Dosher, & Perkins, 1989;Lappin & Fuqua, 1983;Rogers & Graham, 1979;Wallach & O'Connell, 1953).…”