4.inverse Filter Deconvolution2017

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

11/23/2017

Inverse Filter dan Dekonvolusi

TG-4112: Analisa Sinyal Geofisika


1

Dekonvolusi ?
• deconvolution is an algorithm-based process used to reverse the effects
of convolution on recorded data.

• The concept of deconvolution is widely used in the techniques of signal


processing and image processing

• In general, the objective of deconvolution is to find the solution of a


convolution equation of the form:

• f* g =h

• h is some recorded signal, ƒ is some signal that we wish to recover, but has
been convolved with some other signal g before we recorded it.
• The function g might represent the transfer function of an instrument or a
driving force that was applied to a physical system

• Inverse filtering is the simplest method for deconvolution


TG-4112: Analisa Sinyal Geofisika
2

1
11/23/2017

2
11/23/2017

3
11/23/2017

The seismic impulse response

7
Linear systems Computational Geophysics and Data Analysis

The filtered response

8
Linear systems Computational Geophysics and Data Analysis

4
11/23/2017

1D convolutional model of a seismic trace

The seismogram of a layered medium can also be


calculated using a convolutional model ...

u(t) = s(t) * r(t) + n(t)

u(t) seismogram
s(t) source wavelet
r(t) reflectivity

9
Linear systems Computational Geophysics and Data Analysis

Deconvolution

Deconvolution is the inverse operation to convolution.

When is deconvolution useful?

10
Linear systems Computational Geophysics and Data Analysis

5
11/23/2017

Desain LS Filter

TG-4112: Analisa Sinyal Geofisika


11

Inversion of a dipole
Dipole

Desire output
1 0 0 
a   h0  1 
1 0  0
Find a filter of length 3 that inverts the dipole: 0 a 1   h1  ≈
0     0
0 a  h2   
   0
 

In matrix form: Wh≈g

We form an error function to minimize e =|| W h − g ||22


Minimize the error function by taking derivatives with respect
to the unknowns
∇e = 0 = W T Wh − W T g

LS filter h = (W T W)−1 W T g
Actual output gˆ = W h, (gˆ k = w k * hh )

6
11/23/2017

Inversion of a wavelet Wavelet

 w0 0 0
  Desire output
 w1 w0 0
 w2 w1 w0  1 
h0 
   0
 h1  ≈
 w3 w2 w1     0
w4 h2   
Find a filter of length 3 that inverts the wavelet: w3 w2   0
 
 0 w4 w3 

0 0 w4 

In matrix form: Wh≈g

e =|| W h − g ||22

∇e = 0 = W T Wh − W T g

h = (W T W)−1 W T g

gˆ = W h, (gˆ k = w k * h h )

Ideal and Achievable Inversion of a Wavelet:


Pre-whitening

• Trade-off curves and Pre-whitening

Trade-off parameter
Wh≈g

e =|| W h − g ||22 + µ || h ||22

∇e = 0 = W T Wh − W T g + µg

h = (W T W + µ I)−1 W T g  R0 R1 R2 
 
gˆ = W h, (g k = w k * h h ) R = W W =  R1
T
R0 R1 
 
 R2 R1 R0 

Rk : Autocorrelation coefficient
µ = R0 .P /100, P : %of pre − whitening

7
11/23/2017

Ideal and Achievable Inversion of a Wavelet:


Pre-whitening

• Pre-whitening P=0%
wk : Wavelet hk : Filter gˆ k : Actual Output

Time Domain

High Gain at High Frequencies - Noise will be amplified

Frequency Domain
Normalized Amplitude Spectrum

15

Ideal and Achievable Inversion of a Wavelet:


Pre-whitening

• P=10%
wk : Wavelet hk : Filter gˆ k : Actual Output

With Pre-whitening we can control the gain of the inverse filter

16

8
11/23/2017

Ideal and Achievable Inversion of a Wavelet:


Pre-whitening

• P=20%
wk : Wavelet hk : Filter gˆ k : Actual Output

With Pre-whitening we can control the gain of the inverse filter

17

Ideal and Achievable Inversion of a Wavelet:


Pre-whitening & Trade-off curve

e =|| W h − g ||22 + µ || h ||22

= Misfit + µ Model Norm

Resolution -
Misfit Stability +

µ = 1.
* Resolution +
Stability -

µ = 0.1
* µ = 0.01
* Model Norm

18

9
11/23/2017

TG-4112: Analisa Sinyal Geofisika


19

TG-4112: Analisa Sinyal Geofisika


20

10
11/23/2017

TG-4112: Analisa Sinyal Geofisika


21

TG-4112: Analisa Sinyal Geofisika


22

11
11/23/2017

TG-4112: Analisa Sinyal Geofisika


23

TG-4112: Analisa Sinyal Geofisika


24

12
11/23/2017

TG-4112: Analisa Sinyal Geofisika


25

TG-4112: Analisa Sinyal Geofisika


26

13
11/23/2017

TG-4112: Analisa Sinyal Geofisika


27

TG-4112: Analisa Sinyal Geofisika


28

14
11/23/2017

TG-4112: Analisa Sinyal Geofisika


29

TG-4112: Analisa Sinyal Geofisika


30

15
11/23/2017

TG-4112: Analisa Sinyal Geofisika


31

TG-4112: Analisa Sinyal Geofisika


32

16
11/23/2017

TG-4112: Analisa Sinyal Geofisika


33

TG-4112: Analisa Sinyal Geofisika


34

17
11/23/2017

TG-4112: Analisa Sinyal Geofisika


35

TG-4112: Analisa Sinyal Geofisika


36

18
11/23/2017

TG-4112: Analisa Sinyal Geofisika


37

TG-4112: Analisa Sinyal Geofisika


38

19
11/23/2017

TG-4112: Analisa Sinyal Geofisika


39

TG-4112: Analisa Sinyal Geofisika


40

20
11/23/2017

TG-4112: Analisa Sinyal Geofisika


41

TG-4112: Analisa Sinyal Geofisika


42

21
11/23/2017

Kel. Anggota Kelompok Topik

1 Spektral Dekomposisi

2 FX deconvolution

3 Inverse Filter

4 Weiner Filter

5 Color Inversion

6 Spiking Deconvolution

7 Predictive Deconvolution

Inverse Filter dan Dekonvolusi


Ideal Inverse Filter
A(f) F(f) = 1

Finite Length Inverse Filter


A(f) F(f) ≈ 1

Dalam Domain waktu


, ∗ , , ,… , , , ,… , , , )

Dimana A adalah matriks yang berisi dipol, f adalah filter ynag tidak diketahui
(inverse filter), x adalah output aktual dan b adalah output ideal dari inverse filter.
TG-3110: Analisa Sinyal Geofisika
44

22
11/23/2017

Inverse Filter dan Dekonvolusi


Dimana A adalah matriks yang berisi dipol, f adalah filter yang tidak diketahui
(inverse filter), x adalah output aktual dan b adalah output ideal dari inverse filter.
Kita ingin mencari elemen dari filkter f yang errornya minimum yaitu dengan
meminimumkan :

Ini merupakan masalah over-determined least-square dan solusinya adalah :


TG-3110: Analisa Sinyal Geofisika


45

Inverse Filter dan Dekonvolusi



Suku ATA dalam persamaan di atas merupakan matrik autokorelasi R dari dari dipol {a},
Suku kedua adalah koefisien filter yang ingin dicari, suku di sisi kanan adalah korelasi
Silang antara dipol {a} dengan dengan output ideal b.
Ada sejumlah M persamaan linier dengan M variabel yang tidak diketahui fm.
Solusinya adalah

Pendekatan ini dikenal dengan Wiener Filtering, yang dapat menginversi minimum
phase dipol untuk memberikan filter dengan panjang terbatas.
Matriks R memeiliki struktur Toeplitz,

Inverse dari matriks ini dapat dihitung menggunakan algoritma Levinson recursion

TG-3110: Analisa Sinyal Geofisika


46

23
11/23/2017

Inverse Filter dan Dekonvolusi


Stabilisation, damping, regularisation

TG-3110: Analisa Sinyal Geofisika


47

24

You might also like