Abstract
When we recognize a surface material of an object, the scatter distribution from light on its surface can be one of the cues in natural environment light fields (Fleming et al., 2003). We investigated what cues are used when the human visual system is possible to distinguish between "mirror" and "glass" objects under natural and unnatural illumination. Various shaped objects with completely specular reflected surface ("mirror"), and with transparent and refraction components ("glass") were generated by computer graphics. These stimuli were rendered under several real-world (Debevec, 1998) and unnatural illumination composed of binary random noise. Stimuli were presented for 1,000 ms while the object was just standing (static condition) or horizontally rotating (dynamic condition). Human observers were asked to judge its object material (mirror or glass) in a 2AFC paradigm. Under the natural illumination, the performances in the static and the dynamic condition were 75% and 89% showing that observers could perform mirror-glass discrimination even in the static condition and motion information helped the discrimination to some extent. Under the unnatural illumination, however, observers could not judge the object material in the static condition (57%). In the dynamic condition, the performances raised to approximately same level as the performance in the static condition under the natural illumination (78%). We found the similar tendency when observers judged under the natural illumination with color negative/positive inversion in which spatial structure of the reflection remains unchanged. These findings suggest that while static distorted reflection itself does not help to distinguish between mirror and glass under the unnatural illumination, motion information of the distorted reflection of the mirror and glass can be a relevant cue to distinguish them.
Meeting abstract presented at VSS 2016