Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

14
Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3

Transcript of Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

Page 1: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

Image Modeling &

SegmentationAly Farag and Asem Ali

Lecture #3

Page 2: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

2

Parametric methods

These methods are useful when the underlying distribution is known in advance or is simple enough to be modeled by a simple distribution function or a mixture of such functions

The parametric model is very compact (low memory and CPU usage) where only few parameters need to fit.

The model’s parameters are estimated using these methods such as maximum likelihood estimation, Bayesian estimation and expectation maximization.

o A location parameter simply shifts the graph left or right on the horizontal

axis. o A scale parameter (>1) stretches or (<1 ) compress the pdf,

Page 3: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

n

ii

i

n

i

nMLE

p

p

p

1

1

21

)|(logmaxarg

)|(logmaxarg

)|,,,(logmaxarg

x

x

xxx

3

Parametric methods (1- Maximum Likelihood Estimator: MLE)

Suppose that n samples x1, x2, … , xn are drawn independently and identically distributed (i.i.d.) ~ distribution φ(θ) with vector of parameters θ=(θ1,…. , θr)

Know: data samples, the distribution type Unknown : θ ?????

MLE Method estimates θ by maximizing the log likelihood of the data

By i.i.d.

monotonicity of log

To show the dependence of p on ɵ explicitly.

Page 4: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

n

iiMLE

n

iiMLE

n

i

in

i

iT

nn

nLLL

1

22

1

14

2

21

2

)(11

0]2

)(

2,[],[

xx

xx

4

Parametric methods (1- Maximum Likelihood Estimator: MLE)Let

Then calculate

Find θ by letting

n

iipL

1

)|(log x

T

r

LLLL],,,[

21

0L

Example:

Suppose that n samples x1, x2, … , xn are drawn independently and identically distributed (i.i.d.)~ 1D N(μ,σ) find the MLE of μ and σ

n

i

in

i

xn

ii nepL

i

12

2

1

2

)(

1 2

)(

2

1log

2

1log)|(log)(

2

2

x

x

Matlab Demo

In some cases we can find a closed form for θ

Coin Example ………………..

Page 5: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

i

n

ii

n

iiMLE nE

nE

nnEE xxx

111][

11

5

An estimator of a parameter is unbiased if the expected value of the estimate is the same as the true value of the parameters.

Parametric methods (1- Maximum Likelihood Estimator: MLE)

Example:

An estimator of a parameter is biased if the expected value of the estimate is different from the true value of the parameters.

Example:

2

1

2

12

1

2

1

2

11

22

11

............................................

))(12

(1

)1

(1

)(1

][

n

prooftheComplete

nnnE

nnE

nEE

n

i

n

jj

n

ijii

n

i

n

jji

n

iMLEiMLE

xxxx

xxx

n

iMLEiunbiased n 1

22 )(1

1 x doesn’t make much difference once n --> large

Page 6: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

K

kkk xxp

1

;)(

6

What if there are distinct subpopulations in observed data?

Parametric methods

Example Pearson in 1894, tried to model the distribution of the ratio

between measurements of forehead and body length on crabs.

He used a two-component mixture.

It was hypothesized that the two-component structure was related to the possibility of this particular population of crabs evolving into two new subspecies

Mixture Model

The underlying density is assumed to have the form

The weights, Constrained

K

kkk

1

1,0

Components of the mixture are densities and are

parameterized by k

What is the difference between Mixture Model and the kernel-based estimator?

Page 7: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

2

2

1

1 1,12

2

1,11

1

11 n

Ci

i

n

Ci

i

ii

nnxx

7

Parametric methods

Example Given that {xi, Ci1, Ci2} n samples (complete data) drawn i.i.d. two

normal distributions

xi observed value of ith instance

Ci1 and Ci2 indicate which of two normal distributions was used

to generate xi

Cij=1 if Cij was used to generate xi, 0 otherwise

MLE

2

2

1

1 1,1

22

2

22

1,1

21

1

21 )(

1)(

1 n

Ci

i

n

Ci

i

ii

nn xx

How can we estimate the parameters given incomplete data (don’t know Ci1 and Ci2 )?

Page 8: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

8

Parametric methods (2- Expectation Maximization: EM)

The EM algorithm is a general method of finding the maximum-likelihood estimate of the parameters of an underlying distribution from a given data set when the data is incomplete or has missing values.

EM Algorithm:

Given initial parameters Θ0

Repeatedly

o Re-estimating expected values of hidden binary variables Cij

o Then recalculate the MLE of Θ using these expected values for the hidden variables Note:

EM unsupervised method, but MLE supervised

To use EM you must to know:

Number of classes K,

Parametric form of the distribution.

Page 9: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

9

Illustrative example

complete incomplete

Page 10: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

10

Illustrative example

Page 11: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

)|,( cxp

11

Parametric methods (2- Expectation Maximization: EM)

Assume that a joint density function for complete data set

The EM algorithm first finds the expected value of the complete-data log-likelihood with respect to the unknown data C given the observed data x and the current parameter estimates Θi-1.

11 ,|)|,(log)|( ii xcxpEQ

The current parameters estimates that we used to evaluate the expectation

The new parameters that we optimize to maximize Q

The evaluation of this expectation is called the E-step of the algorithm

The second step (the M-step) of the EM algorithm is to maximize the expectation we computed in the first step.

)|(maxarg 1

ii Q

These two steps are repeated as necessary. Each iteration is guaranteed to increase the log likelihood and the algorithm is guaranteed to converge to a local maximum of the likelihood function.

Page 12: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

K

kkk xxp

1

;)(

12

Parametric methods (2- Expectation Maximization: EM)

The mixture-density parameter estimation problem

],,,,,[ 11 KK

dcxcpcxpxcxpEQCc

iii

),|()|,(log,|)|,(log)|( 111

),|( 1

1

i

jj

n

jxcp

K

k

ikj

i

icj

i

ijj

xp

xpxcp

k

jjc

1

1

)|(

)|(),|(

Using Bayes’s rule, we can compute:

),|())|(log()|( 1

1 1

1

ij

K

k

n

jkjk

i xkpxpQ

Then

E-step

Grades Example ………………..

Page 13: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

,

13

Parametric methods (2- Expectation Maximization: EM)

For some distributions, it is possible to get an analytical expressions for For example, if we assume d-dimensional Gaussian component distributions

n

j

ikj

ikj

n

jj

i

xkp

xkpx

k

1

1

1

1

),|(

),|(

n

j

ikj

n

j

Tij

ij

ikj

i

xkp

xxxkpkk

k

1

1

1

1

),|(

))()(,|(

n

j

ikj

i xkpnk

1

1),|(1

M-step

K

k

ikj

i

icj

i

ijj

xp

xpxcp

k

jjc

1

1

)|(

)|(),|(

E-step

Page 14: Image Modeling & Segmentation Aly Farag and Asem Ali Lecture #3.

0 50 100 150 200 2500

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.02

14

Parametric methods (2- Expectation Maximization: EM)

Example:

MATLAB demo