+ All Categories
Home > Education > My research in 2013 in English

My research in 2013 in English

Date post: 06-Jul-2015
Category:
Upload: kazuma-nagafune
View: 198 times
Download: 0 times
Share this document with a friend
25
Genetical Algorithm(GA) for Sparse CT Image Reconstruction. 3.10.2013 Kazuma Nagafune 1
Transcript
Page 1: My research in 2013 in English

Genetical Algorithm(GA) for Sparse CT Image Reconstruction.

3.10.2013

Kazuma Nagafune

1

Page 2: My research in 2013 in English

Background of my study

3.10.2013 2

Before talk of GA

Computed Tomography(CT)

A technique of reconstructing section image.

1. Project to object ALL directions

2. Caluculate for respective throughed quantity

Page 3: My research in 2013 in English

Problems

3.10.2013

IF num of projections are few(Sparse CT) →Can not reconstruct perfectly.

3

Page 4: My research in 2013 in English

Benefits

3.10.2013 4

IF we can solve Sparse CT →Get knowledges some structures of the object.

Page 5: My research in 2013 in English

Arrange problems

3.10.2013

Projection needs ALL directions Obstacle might be there

=Sparse CT

5

Page 6: My research in 2013 in English

Arrange problems

3.10.2013

Projection needs ALL directions Obstacle might be there.

=Sparse CT We want to reconstruct even if our data is lacked.

6

Page 7: My research in 2013 in English

Projection needs ALL directions

Arrange problems

3.10.2013

Projection needs ALL directions Obstacle might be there.

=Sparse CT We want to reconstruct even if our data is lacked.

Approach applying

Genetical Algorithm(GA)

7

Page 8: My research in 2013 in English

fi :=Objective function(i=1,2,…n) := individuals

Genetical Algorithm(GA)

• Optimize Multi objective functions

• Exploitate using multi individuals

3.10.2013 8

Fllow chart f1

f2

Minimize (Maximum)

Objective functions

GS algorithm

Minimum problem

Page 9: My research in 2013 in English

fi :=Objective function(i=1,2,…n) := individuals

Genetical Algorithm(GA)

• Optimize Multi objective functions

• Exploitate using multi individuals

3.10.2013 9

Fllow chart f1

f2

Minimize (Maximum)

Objective functions

GS algorithm

Minimum problem

Page 10: My research in 2013 in English

fi :=Objective function(i=1,2,…n) := individuals

Genetical Algorithm(GA)

• Optimize Multi objective functions

• Exploitate using multi individuals

3.10.2013 10

Fllow chart f1

f2

Minimize (Maximum)

Objective functions

GS algorithm

Minimum problem

Explain the outline of GA

for Real number, NSGA-II[Deb, 2002]

GA is a one of the concenpt of

Evolutionary Multi-criterion Optimization(EMO).

Page 11: My research in 2013 in English

Initialization

• Make individuals

– They are ramdomly.

3.10.2013

Fllow chart

GS algorithm

11

f2

f1

Page 12: My research in 2013 in English

Evaluation and Selection

• Individuals are evaluated and selected

– Related limited conditions

– f1, f2 have different limited conditions

3.10.2013

Fllow chart

GS algorithm

12

f2

f1

Page 13: My research in 2013 in English

GA operator

• Brush up them using GA-operator

3.10.2013

Fllow chart

GS algorithm

13

Parent1 Parent2

Child1 Child2

ああああ Mutation

Page 14: My research in 2013 in English

GA operator

• Genetical Algorithm(GA) operator

– Brush up them using GA-operator

3.10.2013

Fllow chart

GS algorithm

14

Parent1 Parent2

Child1 Child2

ああああ Mutation

In my study, I had cogitated and equipped GA-operator in Japan. (But I’m not going to explain this in this presentation).

Page 15: My research in 2013 in English

Crossover

• Crossover selected individuals(parents) – Set variety feautures to children

from parents’s.

3.10.2013

Fllow chart

GS algorithm

Parent1 Parent2

Child1 Child2

15

Page 16: My research in 2013 in English

Mutation

• Some time mutate a few individuals in the population

– Expect quiet new individual.

3.10.2013

Fllow chart

GS algorithm

16

ああああ Mutation

Page 17: My research in 2013 in English

Evaluation and one aalgorithm

• Evaluate Individuals

• Apply one algorithm

• Checek for terminate

3.10.2013 17

Talk about my case of study considering adobe explanation.

Fllow chart

GS algorithm

Page 18: My research in 2013 in English

My case of study

3.10.2013

Fourier

Transform

Real space

18

Frequency space

• One image has two individuals – Real part and Imaginary part of frequency space)

• If 256 x 256pixels image, one individul has 65536 x 2 genes. – Too much genes will have to be brushed up.

Use all pixels of frequency space for individual’s genes.

Page 19: My research in 2013 in English

Sparse CT is a inverse problem

3.10.2013 19

Lacked data will not know even if GOD.

Given data is very few.

Given data

Lacked data

To trace is quiet difficult.

Because given data is very few.

Page 20: My research in 2013 in English

Sparse CT is a inverse problem

3.10.2013 20

Sparce CT (inverse problem) is inperfect problem,

so observer can not reconstruct original image perfectly.

Original image Reconstructed image

Page 21: My research in 2013 in English

Numerical experience

3.10.2013

・EMO algorithm:for realnumber,NSGA-II

・generation:30

・Original image:Phantom,Interior of watch

・Projections:8(phantom),20(watch)

・Each parameter (Phantom/LENA) :Table 1

Num of GS algorithm

10/20

10/20

21

Original image

Projections

Table 1. each parameter

Page 22: My research in 2013 in English

推定対象画像

Interior a watch

3.10.2013

Phantom

22

Page 23: My research in 2013 in English

Result

3.10.2013

Reconstructed result for 8 projections .

23

Supposed Previous work Original

Page 24: My research in 2013 in English

Result

3.10.2013

Reconstructed result for 20 projections .

24

Supposed Previous work Original

Page 25: My research in 2013 in English

Summarize

The outcomes

• I have conducted on toys I have achieved favorable results.

• I could show the efficacy utilizing GA-operator

Another cahllanges

• Include some noises from given data

• Utilize results another previous work

3.10.2013 25


Recommended