Segmentation of Overlapped Objects

based on IEEE paper titled

Segmentation of Overlapping Elliptical Objects in Silhouette Images

Shabeeb Ali O
No.15

Government Engineering College, Kozhikode

March 2, 2018
Introduction

This project deals with segmentation problem

Image segmentation is a fundamental tool in image processing
for partitioning a digital image into multiple segments
The goal of segmentation is to simplify and/or change the
representation of an image into something that is more
meaningful and easier to analyze.
Objectives

A Silhouette image is given, so that the foreground (objects)

and background can be easily distinguished
Seedpoint Extraction : From the given image, points of
interest is located using morphological methods accompanied
by Fast Radial Symmetry Transform.
Contour Evidence : The points of interest (seeds) are then
used to get the evidence of contour by edge-to-seed
association methods.
Contour Estimation : The obtained contour evidence is then
estimated using ellipse fitting (for circular objects) / object
oriented shape fitting methods.
Steps

Image synthesis : Silhouette

Morphology : Erosion
Fast Radial Symmetry Transform
Contour evidence extraction : Edge-seedpoint association
Contour Estimation :Ellipse fitting
Morphology : Erosion

The field of mathematical morphology contributes a wide

range of operators to image processing, all based around a few
simple mathematical concepts from set theory.
Common usages include edge detection, noise removal, image
enhancement and image segmentation.
Erosion is one of the two basic operators in the area of
mathematical morphology, the other being dilation. It is
typically applied to binary images.
Binary Image: Foreground:white, Background: black
Morphology : Erosion

Erosion Operator takes two arguments

1. Input image to be eroded
2. Set of coordinate points known as a structuring
element(kernel).
It is this structuring element that determines the precise effect
of the erosion on the input image.
Morphology : Erosion

Mathematical Definition
Suppose that X is the set of Euclidean coordinates corresponding
to the input binary image, and that K is the set of coordinates for
the structuring element.
Let Kx denote the translation of K so that its origin is at x.
Then the erosion of X by K is simply the set of all points x such
that Kx is a subset of X.
Morphology : Erosion

For each foreground pixel (which we will call the input pixel)
we superimpose the structuring element on top of the input
image so that the origin of the structuring element coincides
with the input pixel coordinates
Morphology : Erosion
Operation : If for every pixel in the structuring element, the
corresponding pixel in the image underneath is a foreground
pixel, then the input pixel is left as it is. If any of the
corresponding pixels in the image are background, however,
the input pixel is also set to background value.

Figure: Effect of erosion using a 3x3 square structuring element

Morphology : Erosion

Implementation
Erosion is applied on the silhouette image with structuring element
as closed disc B(0,1) of the radius 1 using MATLAB Command:
”strel(’disk’,r,n)” The purpose of erosion process is only to
improve the quality of seedpoint extraction
Fast Radial Symmetry Transform (FRST)
Fast radial symmetry (FRS) transform is a feature extraction
technique that transforms the original image to a new
representation that highlights the local radial symmetry of the
image gradient.
Orientation Projection Image On & Magnitude Projection
Image Mn
These images are generated by examining the gradient g at
each point p from which a corresponding positively-affected
pixel p+ve (p) and negatively-affected pixel p−ve (p) are
determined.
Fast Radial Symmetry Transform (FRST)

The radial symmetry contribution at a range n is defined as

the convolution

Sn = Fn ∗ An

where

Fn (p) = ||O n (p)||(α) M n (p)

On
O n (p) =
maxp {On (p)}

Mn
M n (p) =
maxp {Mn (p)}
α is the radial strictness parameter, and An is a two-dimensional
Gaussian.
Fast Radial Symmetry Transform (FRST)
Demonstration
Contour Evidence Extraction

Edge-to-seedpoint association is the process to be done.

The edge-to-seedpoint association method combines the
distance and the divergence index (cosine distance) to assign
edge pixel points to the seedpoints.
Seed points S = {s1 , s2 , ....sn }
Edge pixel point ek inE = {e1 , e2 , .....em }

1−λ div (ek , sj ) + 1

rel(ek , sj ) = +λ
1 + dist(ek , sj ) 2

where rel() is relevance matrix,dist(), div () are respectively

Euclidean distance and divergence functions
Contour Evidence Extraction

Figure: Contour evidence extraction performed by edge-to-seedpoint

association: (a) Original grayscale image; (b) Gradient image; (c)
Edge-to-seedpoint association
Contour Estimation
After contour evidence collection, edges are obtained for
corresponding seedpoints.
Then contour is estimated for each object represented by
seedpoints, using ellipse fitting method
In algebraic ellipse fitting, an ellipse as a special case of a
generic conic is formulated by the zero set of the second order
polynomial equation.
For a given point (x, y) an ellipse with a parameter vector a is
defined as the equation

a0 x 2 + a1 xy + a2 y 2 + a3 y + a4 y + a5 = 0 (1)

where

a = [a1 , a2 , a3 , a4 , a5 ]
Contour Estimation
Equation (1) defines an ellipse, provided that the quadratic
condition ∆ < 0

∆ = a12 − 4a0 a2 < 0

Reference
[1] Sahar Zafari, Tuomas Eerola, Jouni Sampo, Heikki Kälviäinen,
and Heikki Haario, “Segmentation of Overlapping Elliptical
Objects in Silhouette Images,” IEEE Trans. Image Process.,
vol. 24, no. 12, 2015.
[2] Chiwoo Park, Jianhua Z. Huang, Jim X. Ji and Yu Ding
“Segmentation, Inference, and Classification of Partially
Overlapping Nanoparticles,” IEEE Trans. Pattern Analysis and
Machine Intelligence., vol. 35, no. 3, 2013.
[3] Gareth Loy and Alexander Zelinsky, “A Fast Radial Symmetry
Transform for Detecting Points of Interest,” IEEE Trans.
Pattern Analysis and Machine Intelligence., vol. 25, no. 8, pp.
959-973, 2003.
[4] R. Schettini, G. Ciocca, and S. Zuffi, “A survey of methods for
colour image indexing and retrieval in image databases,” Color
Imaging Science: Exploiting Digital Media, 2001.