Abstract
The projected image of a textured surface contains multiple texture cues to its 3D surface orientation, including the change in the scaling of texture across the image and anisotropic compression of texture due to foreshortening. There remains debate about the role of these cues in 3D slant perception. Some cue perturbation studies have observed a strong influence of texture compression and weaker influence of scaling (e.g., Rosenholtz & Malik, 1997; Knill, 1998), while Todd, Christensen & Guckes (2010) found that slant estimates were predicted by texture scaling but not compression. We investigated the relative influence of texture scaling and compression cues for 3D slant perception using a slant estimation task and stimuli with small cue conflicts. Cue conflict stimuli simulated views of planar surfaces slanted around a horizontal axis by 0°–60°, and covered with texture that was compressed or stretched along the tilt direction by ±10% or ±20%, which alters texture compression but not scaling. Consistent cue stimuli were also presented with slants varying from 0°–70°. We compared slant estimates in cue conflict conditions to those in consistent cue conditions with matching texture compression or scaling cues to assess their relative influence. Experiments 1 and 2 presented monocular images with 10° or 20° field-of-view, with textures composed of either circles or irregular voronoi cells. In all cases, slant estimates were mainly determined by compression information. For the voronoi textures, there were small effects of field-of-view, suggesting an influence of texture scaling, while for the circle textures these effects were not detectable. Experiment 3 presented binocular views of the same simulated conditions and observed an influence of texture compression even with conflicting stereo information. Our results demonstrate that texture compression can substantially influence to perception of slant from texture, consistent with earlier findings that multiple texture cues are utilized by the visual system.
Meeting abstract presented at VSS 2018