Abstract
Recent psychophysical studies utilize adaptation effects to probe the functional representations underlying the perception of complex stimuli such as faces, scenes, and surfaces. Here, we applied an adaptation paradigm for the perception of 3D shape (and material) of a realistic object, and show robust aftereffects for a wide range of attributes. We used computer-generated images of a spherical object that was deformed according to a radial map of band-pass noise of variable frequencies (1-32 cycles/pi) and amplitudes. The deformations at different frequencies resulted in global shape distortion (low), bumpiness (mid), and surface roughness (high). Several material properties, diffuse and specular reflectance and translucency, were also manipulated. In the experiment, the observer first viewed an adapting object for 10 sec. The object was repeatedly flashed for 250 ms at either the left or right side of a fixation point. Different adapting stimuli had particular shape and material parameters; e.g., light, glossy, opaque, and very bumpy. On each trial following top-up adaptation of 3 sec, the test and comparison objects were presented for 250 ms at the adapted and non-adapted location, respectively. The observer compared a particular attribute between the two objects (e.g., which object appears bumpier), and the PSE estimated along a given parameter dimension (e.g., deformation amplitude). It was found that the perceived shape distortion, bumpiness, roughness, lightness, glossiness, and opacity were dramatically reduced following adaptation to an object that had a higher value in that attribute. Adaptation to lower values enhanced the perception of some attributes. The aftereffects were also induced by textures on a matte sphere, or even by a simple band-pass noise image. These results imply an important role of low-level summary statistics not only in material perception, but also in global 3D shape perception.
Meeting abstract presented at VSS 2012