Abstract
Depicted slant is defined as slant based on linear-perspective assumptions about lines imaged on a flat surface. Slants of triangles and trapezoids were computed as a function of depicted slant and slant of the obliquely viewed picture plane. Computations were based on assumptions of parallelism and orthogonality. Perceived slant was measured during binocular viewing in a matching task. The matched slants were compared with computed slants. Matched and computed slants were highly correlated. Residual error analysis showed that both parallelism and orthogonality explained about 95% of the data. Contributions of the picture plane, signaled by binocular disparity and various monocular cues, were small for both trapezoids and triangles. The results imply that the claimed non-Euclidean nature of pictorial space is straightforwardly explained by linear-perspective assumptions. Further analysis indicated that the visual system derives slant from retinal angles alone without requiring knowledge of distance and orientation of the picture plane. Precision of the slant judgments requires a neural substrate that is able to make highly precise comparisons between orientations of lines imaged at different retinal locations. The neural basis of slant from linear perspective has not yet been clarified. Cells with long-range connections in V1, however, have features that suggest an involvement in slant perception.
Meeting abstract presented at VSS 2014