Purchase this article with an account.
Flip Phillips, Roland Fleming; The Veiled Virgin Project: Causal layering of 3D shape. Journal of Vision 2017;17(10):406. https://doi.org/10.1167/17.10.406.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The form of a three dimensional object depends on a number of factors. Nonrigid, laminar materials, such as cloth and skin, take a form that is determined by their physical properties as well those of an underlying object. For example, clothes adopt the global shape of the person wearing them but also exhibit wrinkles, folds, and other features due to the way fabric drapes and self-organizes. Artists and sculptors are intimately aware of this interaction and have learned to depict compelling representations of figures swathed in cloth using solid lumps of marble (e.g., Giovanni Strazza's "Veiled Virgin"). Inspired by such sculptures, we sought to understand how the visual system decomposes a single visible surface relief into distinct causal layers, distinguishing which features are due to the underlying object, and which due to overlying textile. Here we present a series of experiments testing observers' ability to divine the underlying causes of a 3D object's form. We made three irregular landscape surfaces, which we then draped with an opaque lightweight cloth in various configurations. Using 3D scanning we obtained geometric descriptions of the underlying surfaces and the various drapings, which we then used to render ideal opaque Lambertian surfaces of the composite shapes. Using a painting interface presented on an iPad, subjects directly indicated regions of causation on the stimuli (i.e., which ridges appeared to be caused by the underlying landscape, and which by the overlying textile). This allowed us to directly compare the subjects' markings with the geometric features of the stimuli. We show that subjects were strikingly good at separating the underlying structure from the effects of the covering material. This shows that the visual system can separate shape into distinct causes, much like it can decompose image intensities into transparent layers.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only