Abstract
Without traditional luminance/color cues of perceptual transparency, dynamic deformations of a static image can produce a vivid sensation of a transparent layer. Moreover, this subjective transparent layer looks like a liquid when the deformation flow simulates, precisely or crudely, image deformation due to refraction of lights at a flowing liquid surface (Kawabe, et al., 2013, VSS). Transparent liquids not only refract lights from their body, but also reflect light. Refraction and reflection are never independent of each other, since both depend on the surface orientation map. If visual processing properly takes the physics of refraction into account, it is likely to be sensitive to the congruency between the refraction-based image deformation and the specular reflection. To test this, we synthesized, using Blender software, several scenes including transparent liquid flows in which the refraction-based image deformation was congruent with the specular reflection. To make incongruent stimuli, we combined the image deformation taken from one scene with the specular reflection taken from another. Observers viewed the congruent or incongruent scenes, and made ratings as to the strength of liquid and glossiness impression on five point scales. We found both rating scores to be generally high as long as the scenes were dynamic, regardless of the deformation-specular congruency. We also had observers discriminate congruent scenes from incongruent ones, and found that the discrimination was very difficult. These results indicate that, at least when perceiving a dynamic transparent liquid, the visual system is unable to check the congruency between the refraction-based image deformation and the specular reflection. The visual system may heuristically estimate the presence of a dynamic transparent liquid flow from the analysis of characteristic image deformations, likely produced by refraction, independently of analysis of the specular reflections, and without precisely considering the physical liquid structure.
Meeting abstract presented at VSS 2015