Image Processing Reference
In-Depth Information
•
Mapping to two dimensions: the semantically defining side of the LEGO brick
is isolated, which reduces the three dimensional problem to a two dimensional
problem.
•
Stud arrangement: the relative alignment of the studs.
It is reasonable to assume that the baseboard will have a colour different to that of
the brick itself. Baseboard removal then becomes a simple establishment of colour
near the edge of the image, and subtraction of pixels near to that colour. This, how-
ever, still leaves the shadowy edges of the studs, because the shadows have a differ-
ent colour (see Fig. 8.5(a) and (b)).
The background shadows can then be removed using a CA. Define a Von
Neumann-type neighbourhood
n
x
for each cell
x
i
,
j
, such that
n
x
=(
x
N
,
x
S
,
x
W
,
x
E
)
,
where
x
N
=
x
i
−
δ
,
j
,...,
x
i
−
1
,
j
x
S
=
x
i
+
1
,
j
,...,
x
i
+
δ
,
j
x
W
=
x
i
,
j
−
δ
,...,
x
i
,
j
−
1
x
E
=
x
i
,
j
+
1
,...,
x
i
,
j
+
δ
.
That is, the neighbourhood is taken in the usual four directions up to a distance
of
from the current cell, but instead of using a standard Von Neumann neigh-
bourhood, the neighbourhood forms a rectangle of single cell-width instead of the
normal diamond-shape of the Von Neumann neighbourhood. Suppose now that a
background pixel in cell
x
i
,
j
is indicated by
x
i
,
j
=
δ
0. The update rule for the CA can
then be defined as
⊧
⊨
0
,
if
∃
k
N
,
k
S
,
k
W
,
k
E
:
x
i
−
k
N
,
j
(
t
)=
0&
x
i
+
k
S
,
j
(
t
)=
0&
x
i
,
j
(
t
+
1
)=
x
i
,
j
−
k
W
(
t
)=
0&
⊩
x
i
,
j
+
k
E
(
t
)=
0
1
,
otherwise,
Fig. 8.4
A selection of standard LEGO bricks on the left, and some special bricks on the
right
Search WWH ::
Custom Search