Professional Documents
Culture Documents
Dr.G.Malathi
Associate Professor Senior,
School of Computer Science and Engineering
VIT Chennai
Shape Features
• Objects can exhibit various shapes.
• Shape region features or geometric features
characterize the appearance of an object.
• Area: The number of pixels in that object. All
pixels inside that object is refered to as area.
• It includes the object boundary and not the
hole
• Example : P, B
Geometrical features of an object
Shape Features
• Perimeter: Number of pixels present in the
boundary of the object. The number of
foreground pixels that touches the
background.
IMAGE REPRESENTATION
• After segmentation, representing a region
involves two choices.
• Representation of a region in terms of external
characteristics – its boundary-shape.
• Representation in terms of its internal
characteristics – the pixels-regional properties-
color and texture.
• It should be invariant in rotation, translation,
scaling.
Polygonal Approximation of the boundary
0 6 0 6 0 6 0
0 6 0 6 0 6 0
• Redefine the starting point so that the chain
code obtained will have minimum numerical
value.
Shape numbers
• Considering the differential chain code as a
cycle, choose a starting point that gives a
minimum numerical value which is called as
shape numbers which can be used as
descriptors for shape recognition.(boundary
descriptors).
Exercise – Obtain chain code, differential chain code
and shape number for the below image for 4D and 8D
Representation and description
• After the identification of features, the object
must be represented and described.
• Representation involves storing the feature in
a form that can be processed by an image
analysis programs
• After representation the features should be
described using descriptors
Boundary Representation
• The boundaries can be traced and allotted the
respective numbers based on the direction
• The boundary of the object is reduced to a
sequence of numbers
Boundary Representation
Differential chain code
• Normalization solves the problem of the
starting point.
• To normalize the chain code take the first
difference of the chain code by taking two
numbers of the chain code and calculating the
number of transitions required to reach the
second number from the first number in
counter clock wise direction
Boundary Descriptors
• Simple Descriptors: Some of the common methods
to describe boundary are length, diameter
• Boundary length:The number of pixels present in
the boundary gives the boundary length.
• Diameter: Let D(p1,p2) be the line connecting two
points p1,p2 on the boundary. The maximum
distance connecting two points p1 and p2 is called
the diameter
Diameter = max(D(pi,pj))
Boundary Descriptors
• Shape Numbers: Once the normalization is carried out, the
smallest magnitude number is obtained. This number is called
shape number.
• Consider the chain code 1 3 1 0 3 0
• The shape number is obtained by shifting this number by one
position as follows:
• 310301
• 103013
• 030131
• 301310
• 0 1 3 1 0 3 This is the smallest and it is called the shape number
Boundary Descriptors
• Projections: It is a metric that denotes a binary
image as a function of row coordinates The
horizontal projection is defined as