Purchase this article with an account.
Andrea Li, Qasim Zaidi; 3-D shape from chromatic orientation flows. Journal of Vision 2004;4(8):158. doi: https://doi.org/10.1167/4.8.158.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In perspective images of 3-D surfaces, qualitatively veridical shape is conveyed by patterns of orientation flows that occur regardless of the homogeneity of the texture on the surface (Li & Zaidi, 2003). Given that there is a debate about the existence of oriented chromatically sensitive neurons, can purely chromatic flows convey veridical shape? Patterns calculated as isoluminant for the CIE observer can convey qualitatively veridical shape, but for a real observer it is impossible to make an extended image uniformly isoluminant, especially if it is viewed with unrestricted eye movements. We thus used the following procedure: An achromatic texture pattern consisting of the sum of eight gratings of equal frequency and contrast, oriented at 0 (horizontal), ±22.5, ±45, ±67.5 and 90 was repeated along the depth axis to form a volumetric solid that was then carved in depth to form a vertical, sinusoidally curved surface. In the perspective image of this surface, the horizontal grating exhibits the signature orientation flows, but they are not visible as they are masked by the ±22.5 components, and observers perceive non-veridical shapes, even if the contrast of the horizontal component is increased up to 30%. If the achromatic horizontal grating is removed, the remaining pattern does not convey veridical 3-D shape, but if the horizontal component is replaced with an isoluminant L-M grating of the same spatial frequency and orientation, the chromatic orientation flows are readily visible and veridical shape is perceived. Luminance artifacts could account for the shape percept only if they exceeded 30% contrast, which is not physically possible given the chromatic cone contrasts. Thus the orientation flows were detected, and veridical shape perceived, from purely chromatic information. These results show that neural mechanisms used to extract 3-D shape must receive input both from cells tuned to chromatic differences and cells tuned to achromatic differences.
NEI grant EY13312 to Q. Zaidi
This PDF is available to Subscribers Only