Blurring and Local Blurring

Blurring and Local Blurring

姓名 : 張珮毓授課教師：張顧耀 教授報告日期： 2007/06/12

Outline Introduction Blurring

Discrete GaussianBinomial BlurringRecursive Gaussian IIR

Local Blurring

112/04/21

2

Introduction This section describes several methods that

can be applied to reduce noise on images. Blurring

Blurring is the traditional approach for removing noise from images.

The effect of blurring on the image spectrum is to attenuate high spatial frequencies.

112/04/21

3

Outline Introduction Blurring

Discrete GaussianBinomial BlurringRecursive Gaussian IIR

Local Blurring

112/04/21

4

Header: #include "itkDiscreteGaussianImageFilter.h"

filter->SetVariance( gaussianVariance ); filter-

>SetMaximumKernelWidth( maxKernelWidth );

112/04/21

5

Discrete Gaussian

2

2

2

2

1

x

exG

112/04/21

6

Discrete Gaussian

Effect of the DiscreteGaussianImageFilter on a slice from a MRI proton density image of the brain.

112/04/21

7Variance=3 maxKernelWidth=3*3

Discrete Gaussian

Variance=7maxKernelWidth=3*3

112/04/21

8

Discrete Gaussian

Variance=3 maxKernelWidth=5*5

Variance=3 maxKernelWidth=7*7

Binomial Blurring Binomial Blurring computes a nearest

neighbor average along each dimension The process is repeated a number of times, as

specified by the user.

112/04/21

9

Binomial filters. 1D versions

112/04/21

10

Binomial Blurring

112

11b

11 bbb *1214

12

1111 bbbbb ***1464116

14

111111 bbbbbbb *****161520156164

16

In principle, after a large number of iterations the result will approach the convolution with a Gaussian.

Header: #include "itkBinomialBlurImageFilter.h"

112/04/21

11

Binomial Blurring

112/04/21

12

Binomial Blurring

Repetitions=2 Repetitions=8

Finite Impulse Response (FIR) Infinite Impulse Response (IIR)

112/04/21

13

Recursive Gaussian IIR

只延伸到有限距離 (FIR) 延伸到影像邊界 (IIR)

Recursive Gaussian IIR Header:

#include "itkRecursiveGaussianImageFilter.h”

112/04/21

143 5

112/04/21

15

Recursive Gaussian IIR

Sigma=3 Sigma=5 Sigma=7

Outline Introduction Blurring

Discrete GaussianBinomial BlurringRecursive Gaussian IIR

Local Blurring

112/04/21

16

Local Blurring In some cases it is desirable to compute

smoothing in restricted regions of the image, or to do it using different parameters that are computed locally.

112/04/21

17

ItkSoftwareGuide http://www.engr.udayton.edu/faculty/jloomis/

ece563/notes/filter/smooth/smooth.html http://ssip2003.info.uvt.ro/lectures/gui/

smoothing_techniques_in_image_processing.ppt

http://www.cee.hw.ac.uk/hipr/html/gsmooth.html

112/04/21

18

參考資料