Scribd
Upload
21 views

Course 18.327 and 1.130 Wavelets and Filter Banks

1. The document discusses various methods for handling finite-length signals in wavelet analysis and filter banks, including zero-padding, periodic extension, and symmetric extension. 2. Symmetric extension ensures continuity of function values at boundaries but not derivatives. It can be whole-point symmetric or half-point symmetric depending on the filter. 3. Polynomial extrapolation provides another method to extend signals by fitting polynomials and allows incorporating boundary conditions from partial differential equations.

Uploaded by

djoseph_1
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
21 views

Course 18.327 and 1.130 Wavelets and Filter Banks

1. The document discusses various methods for handling finite-length signals in wavelet analysis and filter banks, including zero-padding, periodic extension, and symmetric extension. 2. Symmetric extension ensures continuity of function values at boundaries but not derivatives. It can be whole-point symmetric or half-point symmetric depending on the filter. 3. Polynomial extrapolation provides another method to extend signals by fitting polynomials and allows incorporating boundary conditions from partial differential equations.

Uploaded by

djoseph_1
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Course 18.327 and 1.

130 Wavelets and Filter Banks


Signal and Image Processing: finite length signals; boundary filters and boundary wavelets; wavelet compression algorithms.

Finite-Length Signals

x[n]
y-2 y-1 y0 y1  yN-1 

H()

y[n]
x-2 x-1 x0 x1  xN-1 
unknown

h0 h1 h0 h-1 h2 h1 h0 h-1 h1 h0 h-1   

length N (finite length) unknown

1) zero-padding

x[n] n

 -2 -1 0 1 2 3  filtered by [1, 1] y[n]  -2 -1 0 1 2 3 filtered by [1, -1] y[n]  -2 -1 0 1 2 3

artificial edge resulting from zero-padding

2) Periodic Extension - 1 0 1 2 N-1 N y0 y1  yN-1 =

x[n] n

x[N] = x[0]

wrap-around

h1 h0 h-1 h-2  h2 h1 h1 h0 h-1  h3 h2 h-1 h-2  h2 h1 h0

 xN-2 xN-1 x0 x1 x2  xN-1 x0 

N-output

circulant matrix = H
4

What is the eigenvector for the circulant matrix? [ 1 ei ei2  ei(N-1) ] T We need eiN = 1 = ei0 N = 2k For the 0th row,
N-1
2 k

2k N

discrete set of s H[k] = h[n] e-i N n


n=0

1 1 1  1 w w2 [H] 1 w2 w4    1 wN-1 k=0k=1 F HF = F H[k] = h[n]e-i


n n 
2 k Nn

1 H[0] N-1 H[1] w  w2(N-1) = [F]  H[N-1] k=N-1 w = ei N


2

contains the Fourier coefficients


2 k Nn

h[n - ]x[]e-i If x[] = ei


2 k0 N 

= H[k]X[k]

X[k] = [k k0] H[k]X[k] = H[k0]X[k]

3) Symmetric Extension 1) Whole point symmetry when filter is whole point symmetric. 2) Half point symmetry when filter is half point symmetric. e.g. Whole point symmetry: filter and signal h0 h1 h1x2 + h0x1 + h1x0 h1x1 + h0x0 + h1x1 h1 h0 h1 = h1x0 + h0x1 + h1x2 h1 h0 h1
  

x2 x1 x0 x1 x2

e.g. whole point symmetry filter, half-point symmetry - signal h1x2 + h1x1 + h1x0 + h1x0 + h0x1 + h1x0 h0x0 + h1x0 h0x0 + h1x1 h0x1 + h1x2 h1 h0 h1 h1 h0 h1 = h1 h0 h1
  

x2 x1 x0 x0 x1 x2

Half point symmetry

Whole point symmetry

Downsampling a whole-point symmetric signal with even length N at the left boundary: x
-2 -1 0

x
1 2

still whole-point symmetric after 2.

at the right boundary: x x N-1 odd E.g. 9/7 filter: whole-point symmetric use the above extension for signal N N/2 exactly N/2
9

half-point symmetric after 2.

Downsample a half-point symmetric signal x x 2 nothing guaranteed

x -3 -2 -1 0 1

Linear-phase filters H() = A()e-i 1) half-point symmetric, 2) whole-point symmetric, = fraction = integer

Symmetric extension of finite-length signal X() = B()e-i


10

The output: Y() = H()X() W W H H W H H W W H W H

W = whole-point symmetry H = half-point symmetry

The above extensions ensure the continuity of function values at boundaries, but not the continuity of derivatives at boundaries.

11

4) Polynomial Extrapolation (not useful in image processing) Useful for PDE with boundary conditions.
x0 x1 x2 x3

4 coefficients fits up to 3rd order polynomials. 2 3 a + bn + cn2 + dn3 = x(n) 1 1 1 1 0 1 2 3 0 0 1 1 22 23 32 33 A


12

0 1

a b c d

x0 x1 x2 x3

Then, x1 = [1 -1 1 -1]

a b = [1 -1 1 -1] A1 c d

x0 x1 x2 x3

PDE f(x) = ck(x k)


k

Assume f(x) has polynomial behavior near boundaries


p-1 i=0

ixi = f(x) = ck(x k)


k

{ ( - k)} orthonormal i (x k)xidx = ck


i=0 p-1

ik
13

0 0 0 1 

1  p-1 0 0 1 2 1 1 

0 1  p-1 =

c0 c1  cp-1

Using the computed is, we can extrapolate, 0 0 1 p-1 e.g. c1 = [1 1  1 ]  p-1 DCT idea of symmetric extension cf. DFT X[k] = x[n]e-i
n
2k n N

complex-valued

Want real-valued results.

14

1 
2 k

N-1

2N-1

2N

DFT of this extended signal:


N-1 n=0

x[n]ei 2N n + x[2N-1-n]e-i 2N n
n=N N-1 N-1
2 k

2N-1

2 k

m=0
2 k

x[m]e-i 2N
2 k

(2N-1-m)

= x[n] {e-i 2N n + e-i 2N (2N-1-n)}


n=0

X(k)

2 x[n]cos ck N

N-1

n=0

k (n+) N
1

 DCT II used in JPEG


k=0 k = 1, 2, , N - 1 15

ck =

You might also like