Image Descriptors

BY: PROF. JYOTSNA SINGH

Descriptors
After performing image segmentation, a region may be represented in
terms of
–external characteristics (boundaries).
–internal characteristics (texture).

A shape may be considered as a filled region with a unique value,

e.g. f (x,y)=1.
Introduction
• The objective is to represent and describe the resulting aggregate of
segmented pixels in a form suitable for further computer processing
after segmenting an image into regions.
• Two choices for representing a region: ‰
➢ External characteristics: An external representation is chosen
when the primary focus is on shape characteristics. e.g. its
boundary; ‰
➢ Internal characteristics: is selected when the primary focus is on
reflectivity properties, such as color and texture..
• For example, A region may be represented by
◦ (a) its boundary described by features such as its length,
◦ (b) the orientation of the straight line joining the extreme points, and
◦ (c) the number of concavities in the boundary.
Image representation
1. Representation schemes
1.1 Chain codes
1.2 Polygonal approximations
1.3 The skeleton of a region
1.4 Boundary descriptors
2. Some simple descriptors
2.1 Shape numbers
2.2 Fourier descriptors
2.3 Moments
3. Region descriptors
3.1 Some simple descriptors
3.2Texture descriptors
1. Representation schemes
• The segmentation techniques yield raw data in the
form of pixels along a boundary or pixels contained in
a region.
• Although these data are sometimes used directly to
obtain descriptors, standard practice is to use schemes
that compact the data into representations that are
considerably more useful in the computation of
descriptors.
• This section introduces some basic representation
schemes for this purpose.
1.1 Chain Code
•To represent a boundary by a connected sequence of straight line
segments of specified length and direction.
• The direction of each segment is coded by using a numbering
scheme
a) Four directional numbering scheme
b) Eight directional numbering scheme

This method generally is unacceptable to apply for the chain codes

to pixels:
(a) The resulting chain of codes usually is quite long;
(b) Sensitive to noise: any small disturbances along the boundary
owing to noise or imperfect segmentation cause changes in the code
that may not necessarily be related to the shape of the boundary.
Chain Code
Chain code is a lossless compression technique used for representing an
object in images.
The co-ordinates of any continuous boundary of an object can be
represented as a string of numbers where each number represents a
particular direction in which the next point on the connected line is
present.
One point is taken as the reference/starting point and on plotting the
points generated from the chain, the original figure can be re-drawn.
Chain Codes
Chain Code
 Chain code for 3 D image
The chain codes could be generated by using conditional statements for each
direction but it becomes very tedious to describe for systems having large
number of directions(3-D grids can have up to 26 directions). Instead we use a
hash function.
The difference in X(dx) and Y(dy) co-ordinates of two successive points are
calculated and hashed to generate the key for the chain code between the two
points. dx dy C(dx, dy)
chainCode[C]
1 0 5 0
Chain code 1 1 8 1
list: 0 1 7 2
Hash function: -1 1 6 3
-1 0 3 4
The function does not generate the value 4 so a dummy value -1 -1 0 5
is stored there. 0 -1 1 6
1 -1 2 7
Chain Code
Chain Code
1. If we change the start point , reflex and rotate the image ----
whole chain code changes
2. Solution is normalised chain code

Steps to Formalise normalise chain code:

1. Calculate the Chain
2. Calculate the Chain first difference
3. Calculate normalise chain code/ Shape Number
 Normalised Chain code
Normalization for starting point:
Treat the code as a circular sequence and
redefine the starting point s that the
resulting sequence of numbers forms an
integer of minimum magnitude. •

Normalization for rotation: Use the first

difference of the chain code instead of the
code itself. The difference is simply by
counting (counter-clockwise) the number of
directions that separate two adjacent
elements of the code.
Example: The first difference of the 4-
direction chain code 10103322 is 33133030.
Normalization for size: Alter the size of the
resampling grid.
Circular First difference
Normalised Chain Code
1. Find “0” in string and then write all numbers coming after 0
including 0. (030)
2. then write the remaining numbers as it after this. (0303133)
this is NCC
3. The normalised chain code is invariant to change in starting
point, reflex and rotation.
Invariance to Reflex and
rotation
 1.2 Polygonal approximation
• The objective is to capture the essence of
the boundary shape with the fewest possible
polygonal segments.
• This problem in general is not trivial and
can quickly turn into a time-consuming
iterative search.
(a) Minimum-perimeter Polygons
◦ A given boundary is enclosed by cells.
We can visualize this enclosure as
consisting of two walls corresponding to
the inside and outside boundaries of the
cells. If the boundary is a rubber band, it
will shrink and take the shape as in (b).
◦ The error in each cell would be at most 2
d , where d is the pixel distance.
 Polygonal approximation
(b) Merging Technique (c) Splitting Techniques
• It is based on error or other • To subdivide a segment successively into
criteria have been two parts until a given criterion is
applied to the problem of satisfied.
polygonal approximation. • Example: a line a drawn between two
• One approach is to merge end points of a boundary. The
points along a boundary until the perpendicular distance from the
least square error line fit of the line to the boundary must not exceed a
points merged so far exceeds a preset threshold. If it does, the farthest
preset threshold. points becomes a vertex.
• Vertices do not corresponding
to corners in the
boundary.
 1.3 The skeleton of a region
(a)Medial axis transformation
•The structural shape of a
plane region can be reduced
to a graph. • The skeleton of a region may be defined
• This reduction can be via the
accomplished by obtaining medial axis transformation (MAT)
the skeleton of the region via proposed by Blum in 1967.
a thinning algorithm.
• Given a region R and a border B:
➢For each point p in R, we find its
closest neighbor in B. If p has more
that one such neighbor, it belongs to
the medial axis (skeleton of R).
➢The concept of "closest" depends on
the definition
of a distance.
 1.3 The skeleton of a region
(b) Thinning algorithm for binary regions
• Assume region points have value 1 and
background points 0.
• A contour point is any pixel with value 1
and having at least one 8-neighbor valued 0.
• The thinning method consists of successive
passes of two steps applied to the contour
points.
1.3 The skeleton of a region
1.3 The skeleton of a region
 2 Some simple descriptors
2.1 Shape numbers
• Shape number of a
A. • Area of the object
boundary is defined as
• Length of a contour the first difference of a
chain code of the
• Diameter of a boundary:
smallest
The maximum distance
between any 2 points on magnitude.
the boundary. • The order n of a shape
number is the number of
digits in its
representation.
• The following figures
shows all shapes of order
4 and 6 in a 4-directional
chain code:
Boundary following
Several algorithms require the points in an ordered
clockwise (or counterclockwise) direction.
• We will describe an algorithm whose output is an ordered
sequence of points.

◦ Binary images (object and background points).

◦ Images are padded with a border of zeros to avoid object
merging with the image borders.
◦ We limit the discussion to single regions.
◦ -The extension is straightforward.
Boundary following
Boundary following
Boundary following
Boundary following
Boundary following
 2 Some simple descriptors
2.2 Fourier Descriptors: In general, only the first few
coefficients are of
Coordinate pairs of points significant magnitude and are
encountered in traversing an N-point pretty enough to
boundary in the xy plane are
recorded as a sequence of complex describe the general shape of
numbers. the boundary.
Example : {(1,2), (2,3),
(2,4),..(x,y),..} • Fourier descriptors are not
directly insensitive to
 {1+2i, 2+3i, 2+4i,..x+yi,..}.
geometrical changes such as
• An N-point DFT is performed to translation, rotation and scale
the sequence and the complex changes, but the changes can
coefficients obtained are called the
Fourier descriptors of the boundary. be related to simple
transformations on the
descriptors.
 2 Some simple descriptors

2.2 Fourier Descriptors:

2 Some simple descriptors
2.3 Moments
• Coordinate pairs of points encountered
in traversing an N-point boundary in the
xy plane are recorded as a sequence of
complex numbers {g(i):i=1,..N}.
Example : {(1,2), (2,3), (2,4),..(x,y),..} 
{1+2i, 2+3i, 2+4i,..x+yi,..}.
• Normalize the area of the object to unit
area.
• The moments of the boundary is given
as
 3. Region Descriptors
3.1 Some simple descriptors
• The area of a region is defined as the
number of pixels contained within its
boundary.
• The perimeter of a region is the length of
its boundary
3.2 Texture descriptors
• Descriptors providing measures of
properties such as smoothness, coarseness
and regularity are used to quantify the
texture content of an object.