# CH.Venkanna EC104018 M.Tech (E.

I)
1

Introduction to image segmentation
y The purpose of image segmentation is to partition an

image into meaningful regions with respect to a particular application y The segmentation is based on measurements taken from the image and might be greylevel, colour, texture, depth or motion

2

Introduction to image segmentation
y Usually image segmentation is an initial and vital

step in a series of processes aimed at overall image understanding y Applications of image segmentation include
y Identifying objects in a scene for object-based

measurements such as size and shape y Identifying objects in a moving scene for object-based video compression (MPEG4)

4

Introduction to image segmentation
y Example 1 y Segmentation based on greyscale y Very simple model of greyscale leads to inaccuracies in object labelling

5

Different Methods of image segmentation
y There are different types of segmentation methods are

proposed. These are
y Thresholding y Clustering

y Relaxation labelling y EM algorithm(Proposed Algorithm)

The Expectation/Maximization (EM) algorithm
y In general we may imagine that an image comprises L

segments (labels)
y

y

Within segment l the pixels (feature vectors) have a probability distribution represented by pl ( x | U l ) U l represents the parameters of the data in segment l y Mean and variance of the greylevels y Mean vector and covariance matrix of the colours y Texture parameters

7

The Expectation/Maximization (EM) algorithm
y The incomplete data are just the measured pixel greylevels or feature vectors y We can define a probability distribution of the incomplete data as pi ( x;U1 ,U 2 .....U L ) y The complete data are the measured greylevels or feature vectors plus a mapping function f (.) which indicates the labelling of each pixel y Given the complete data (pixels plus labels) we can easily work out estimates of the parameters l : l ! ..L y But from the incomplete data no closed form solution exists

The Expectation/Maximization (EM) algorithm
y The algorithm is as follows:

Initialize an estimate of l : l ! 1..L Repeat Step 1: (E step) Obtain an estimate of the labels based on the current parameter estimates Step 2: (M step) Update the parameter estimates based on the current labelling Until Convergence
9

The Expectation/Maximization (EM) algorithm
y A recent approach to applying EM to image

segmentation is to assume the image pixels or feature vectors follow a mixture model
y Generally we assume that each component of the

mixture model is a Gaussian y A Gaussian mixture model (GMM)
l 1

pl ( x | U l ) !

1 1 exp( ( x  Ql )T 7l1 ( x  Ql )) (2T ) d / 2 det(7 l )1/ 2 2

§

L

l !1

El ! 1
10

p ( x | 5)

§

L

E l pl ( x | U l )

The Expectation/Maximization (EM) algorithm
y Our parameter space for our distribution now includes

the mean vectors and covariance matrices for each component in the mixture plus the mixing weights

5 ! Q , 7 , E ,..........Q L , 7 L , E L
y We choose a Gaussian for each component because the

ML estimate of each parameter in the E-step becomes linear

11

The Expectation/Maximization (EM) algorithm
y Define a posterior probability P(l | x j ,U l ) as the

probability that pixel j belongs to region l given the value of the feature vector x j y Using Bayes rule we can write the following equation
P (l | x j , U l ) !

E l pl ( x j | U l )

§

L k !1

E k pk ( x | U k )

y This actually is the E-step of our EM algorithm as

allows us to assign probabilities to each label at each pixel
12

The Expectation/Maximization (EM) algorithm
y The M step simply updates the parameter estimates

using MLL estimation
E
( m 1) l

1 n ! § P(l | x j U l( m ) ) n j !1
n

x j P (l | x j , U l( m ) ) §

Ql( m 1) !

j !1 n

P (l | x j , U l( m ) ) §
j !1

§ P (l |
7 l( m 1) !
j !1

n

, U l( m ) ) ( j
n

_ 

Ql( m ) )( j
, U l( m ) ) j 

Ql( m ) )T a j

§ P(l |
j !1

13

Fig : Original image

Fig : (Right column) The proposed image segmentation, (left column) direct segmentation

Extraction of object from an image
y After partition an image into meaningful

regions and sometimes it is required extract object from an image y Extraction founds many applications in Biometric image processing and face recognition. y Face Feature Extraction(voronoi Diagram method) 
Detecting eyes  Detecting nose and mouth

Fig : Extraction of eyes ,nose and mouth from an image

Extraction of alphabet from number plate

Fig : Original image and segment image,extracting alphabet .

Conclusion
y The EM algorithm is general image segmentation

technique in computer vision with many applications. y Extraction found many applications in security system and biometric image processing.

19