99
2.2.- What is a shape?
A region is simply a connected portion of a plane limited
by a curve boundary. That is, no s e I f- i n te rs e c t i n g
boundary. It is a closed boundary.
A given region has a size, a position and an orientation
in the plane. This defines a flat region, which is defined
uniauely by the curve it has as its boundary. This section
deals with shapes of regions.
A shape is w h a t remains of a region after disregarding its
size, position and orientation in the plane. That is, two
regions have the same shape if we can make them coincide
exactly by translation and rotation intheplane, in addition
to a uniform change of scale (the x and y co-ordinates
increase by the same factor).
A region and its mirror images will not have the same
shape, in general.
This definition coincides with the intuitive
psychological definition of "shape".
If a notation is going to be used to represent the shape
of a region, it has to be independent of the position,
orientation and size of such region . It shouldbe
reproducible: a region, when translated, magnified and
rotated should still give the same description as when
u n tra ns f o rm e d . Two regions with different shapes should
produce different descriptions. Finally, the snape number
should be unique for a given region; for instance, it should
not depend on an arbitrary starting point or a particular
co-ordinate system,
If the notation can be deduced exclusively from the
region, without comparison with a table of canonical shapes
or shape descriptors, for instance, then we can expect
savings in memory and computer time for the procedure that
computes the shape description .
2.3. Continuous and discrete shapes.
A shape is discrete IT FFTe boundary oT The region is formed by
segments of a square grid. Otherwise, the shape is
continuous.
2.4. Matching a continuous shape into a discrete shape.
A Squar e grid may be overraid o n top o7 a continuo us shape
to obtain a discrete shape. The quantization of the shape
is as follow: a square of the grid is "black" (inside the
discrete shape) if more than 50% of it is covered by the
continuous shape; otherwise it is "white" or outside. The
size, orientation and position of this grid will influence
the resulting discrete shape.
A discrete shape obtained from a continuous shape in
the above manner, can not be a shape descriptor of the
continuous shape, because it depends on the size and
orientation of the grid. This will be solved in Section 2.7.
Now, some shape descriptors will be given.
2.5. Eccentricity.
The e c centricity (ra tio of the major to minor axis) of a
region is a descriptor that depends only on its shape .
The major axis of a region is the line joining the two
perimeter points farthest away from each other. The minor
axis is perpendicular to the major axis, and of length sucTT