Item

Abstract

Many commonly occurring substances are somewhat translucent (e.g. wax, jade, fruit-flesh, and cheese). When light strikes a translucent material, it passes through the surface and scatters
a number of times within the body of the object before re-emerging. This causes light to
‘bleed’ through translucent objects, giving them a distinctive visual softness and glow. What
image cues are responsible for this characteristic appearance? How do we distinguish between
translucent and opaque materials? Here we use a combination of image statistics and psychophysics
to study the perception of translucent materials.
There has been a large amount of previous work on the perception of materials that transmit
light. Almost all of this work is based on simple physical models of transparency, (e.g. the
episcotister model of Metelli, 1974), in which the object of interest is a thin filter or screen.
However, recent advances in computer graphics (Jensen et al. 2001; Jensen and Buhler, 2002)
make it possible to simulate the complex physics of solid translucent objects. We have used this
model to study how a wide range of factors influence the perception of translucency, including
highlights, colour, contrast, brightness, blurriness, and conditions of illumination.
Our main findings are as follows: (1) We have found that it is possible to enjoy a vivid
impression of translucency even when many of the cues that were traditionally believed to be
important for the perception of transparency (e.g. X-Junctions, contrast conditions) are absent
from the display. (2) We argue that sub-surface light scattering is too complex for the visual
system to infer translucency by inverse optics. Accordingly we suggest that the visual system
tracks low-level image statistics that reliably correlate with changes in translucency. (3)We find
that perceived translucency varies dramatically with conditions of illumination. We compare
how well a number of candidate cues can predict these variations.
In conclusion, there is a wide range of materials that have not been studied before, and
which we are only just beginning to understand. Many intuitions that we have about which
cues are crucial for recovering opacity turn out to be at best incomplete.