Lines%20and%20Shapes-from-shadows

1
Lines and Shapes-from-shadows

• Juan Bai

Purpose Where do lines fail in vision? What is
its implication for the function of lines in
touch?
2
Lines and Surfaces
We see objects by means of their surfaces, and
source of illumination.
We touch objects by means of their surfaces,
i.e., no brightness information is involved.
For example, touch can often accurately discern
this line drawing as depicting a cup.
3
Lines, Surfaces, and Shadows
Lines copying surface edges give us surface edge
impressions (e.g., roof, hill, etc. in the image
below).
However, lines copying shadow borders do not give
us the impression of the darkness of shadow
(e.g., the shadow on the ground from Kennedy,
1993).
4
Contour, Line, and Outline
Contour reflectance/emission change on a
surface.
A line has two contours, one on each side.
Lines copying surface edges can be called
outlines.
The image below shows that outlines cannot show
shadows (Kennedy, 1993).
5
Herings Outlined Shadow
penumbra
umbra
b
a
Hering (1874) might contend that for the cups
cast shadow in (a), shadow perception fails in
(b) because the penumbra is lost to the dark line
(after Goldstein, 2002).
6
Outlined Sharp-contoured Shadow
a
b
But shadows without penumbras (a, adapted from
Mooney, 1957) convey information about a girls
face, and dark lines in (b) impair percept of the
same face.
Note that shadowed areas in (b) are darker than
illuminated areas (Cavanagh Leclerc, 1989), yet
shape-from-shadow perception is diminished.
7
Shape-from-Shadow Figure-Ground?
Here is another shape-from-shadowed face (adapted
from Mooney, 1957), shown with a color image.
8
Lines, Contours, and Kinds of Borders
Here are different kinds of figure-ground
segregations along borders (Kennedy, Juricevic,
Bai, 2003)
9
Theory of Dark Lines and Shadows
d
b
a
c
Line version (a), negative (c), and
Hering-dark-line version (d) all have negative
contours (light-to-dark from shadow to
non-shadow), and all three fail to show the face
in (b).
So it may be the dark lines negative contour
bordering the shadow in (d) that diminishes
shape-from-shadow perception a
border-polarity hypothesis.
10
Alternative Theory
c
a
b

• Alternatively, dark lines in (a) and (b) may
  impair shape-from-shadow perception because a
  line has two contours rather than one (c) a
  number-of-contours account.

The number-of-contours account is based on an
axis theory of a line (Kennedy, 1993).
11
Kennedy Bai (2000) Polarity vs.
Number-of-contours
a
c

b
Border-polarity predicts the face in (a) should
be easy to see, but number-of-contours predicts
otherwise.
Results (a) and (c) were equally easy, only (b)
difficult, supporting border-polarity.
12
An arm-and-hand is relatively easy to see in
gray-line version (a) but not in black-line
version (b). It is evident in
no-line version (c). So only dark line fails.
Demonstrations for K B (2000)
c
a
b
arm-and-hand (adapted from Kennedy, 1988)
13
Kennedy Bai (2004) Testing Belongingness
c
a
b
A belongingness hypothesis may propose that a
dark-to-light shadow border has to belong to both
the shadowed and the illuminated areas to allow
perception of a continuous face bearing the
shadow. The purpose of K B (2004) is to test
this hypothesis.
14
K B (2004)s Dotted Shadow for Belongingness
Test
Result the dot-grating figure showed the face,
possibly because on average the border polarity
is positive, despite negative polarity along the
final dots.
15
When two areas share a common border, it is said
that the common border belongs to the closer area
the figure it does not belong to the area that
is farther away the ground (e.g., Nakayama et
al., 1989).
Rationale for K B (2004) Stereo Test
If so, when stereo depth indicates that the
bordering elements belong only to a closer,
shadow area or a closer, non-shadow area, the
inappropriate belongingness should impair
shape-from-shadow perception.
16
Only right eye sees smaller
dots
Stereo Stimuli
unpaired dots
most shadow
17
Only left eye sees smaller dots
Stereo Stimuli (Continued)
most shadow
unpaired dots
18
Both eyes see smaller dots at the shadow
border
Stereo Stimuli (Control)
19
Stereo Stimuli (Negative Control)
Results in binocular percept, positive images
showed the face in shadow, even when stereo depth
signaled different belongingness of the border.
Negatives did not show the face.
The results rejected predictions from
belongingness, but supported border-polarity.
20
Conclusions
A Hering line added along a shadow border may
have impaired shape-from-shadow perception
because of negative polarity of its contour
bordering the shadow.
There are 8 kinds of borders in a visual display
that can cause perception of surfaces. These
include luminance/spectral, monocular/binocular,
and static/moving borders (Kennedy et al., 2003).
Among these 8 combinations, only luminance
matters for shape-from-shadow From shadow to
non-shadow the border has to be from darker to
lighter to allow perception of shapes from
shadows.