Вы находитесь на странице: 1из 20

Wavelets and multiresolution

Lecture 16 Milan Gavrilovic milan@cb.uu.se


Centre for Image Analysis Uppsala University

Computer Assisted Image Analysis 2008-05-22

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

1 / 20

Today

Today
Stationary and non-stationary signals. Short-time Fourier transform. Wavelet transform. Multiresolution processing.
Image pyramids. Subband coding.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

2 / 20

Constant Spectra
Stationary signal x (t) = cos (5t 2) + cos (10t 2) + cos (20t 2) + cos (50t 2) Signal in spatial domain Fourier transform of signal

(Images taken from


http://users.rowan.edu/polikar/WAVELETS/WTtutorial.html)
M. Gavrilovic (Uppsala University) L16 Wavelets and multiresolution 2008-05-22 3 / 20

Time Varying Spectra


Non-stationary signal Signal in spatial domain Fourier transform of signal

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

4 / 20

Frequency Components
Stationary signal Non-stationary signal

Both signals constitute of the same frequency components. The signals are undistinguishable in the Fourier domain. Fourier transform is not suitable for analysis of non-stationary signals.
M. Gavrilovic (Uppsala University) L16 Wavelets and multiresolution 2008-05-22 5 / 20

Spectral and Temporal Information

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

6 / 20

Short-Time Fourier Transform

Assume some portion of non-stationary signal as stationary. Divide signal into small segments, such that stationary assumption holds. Choose windowing function w of width equal to signal segment. Position window at t = 0, multiply signal with window, then Fourier transform as with any other signal (t - new information). Position window at t = . . . a.s.o. one Fourier transform for each time step.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

7 / 20

Short-Time Fourier Transform


Short-Time Fourier Transform (STFT)
STFTX
()

t ,f =
t

x (t) t t

ej2ft dt

STFT of signal FT of signal multiplied by window function. Non-stationary signal STFT spectra

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

8 / 20

Drawback with Short-Time Fourier Transform


Window of nite length, covers only portion of signal.
Narrow window Good time res, bad freq res. Wide window Good freq res, bad time res. Heisenbergs uncertainty principle.

Example of using a wide window.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

9 / 20

Wavelet Transform

Continuous Wavelet Transform


CWTx (, s) = (, s) = x

1 |s|

x (t)

t s

dt

Function of two variables:


- translation. s - scale (frequency1 ).

(t) is the transform function, and is called the mother wavelet.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

10 / 20

Wavelet Transform

Non-stationary input signal.

Wavelet transform.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

11 / 20

Varied Scale Shifting


High frequency (small scale). Low frequency (large scale).

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

12 / 20

Image Pyramid and System Block Diagram

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

13 / 20

Image Pyramid and System Block Diagram


Steps

Compute reduced-resolution approximation of input image. Done by ltering (approximation lter) and downsampling. Upsample output of (1) and interpolate. This creates a prediction image. Compute the differences between the prediction in (2) and the input image in (1). This difference is called the prediction residual and can later be used to reconstruct the original image.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

14 / 20

Image Pyramids and Statistics

Gaussian pyramid (Gaussian approximation lter). Laplacian pyramid (prediction residual).

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

15 / 20

Subband Coding
One dimension

Image is decomposed into a set of bandlimited components (subbands). Two-band lter bank for one-dimensional subband coding and decoding. Its spectrum splitting properties. Filter banks satisfying (n) = x(n) are called x biorthogonal.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

16 / 20

Subband Coding
Two dimensions

Two-dimensional four-band lter bank for subband image coding.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

17 / 20

Discrete Wavelet Transform (DFT)


Haar Basis Functions

The basis functions of the Haar transform are the simplest known orthonormal wavelets. Used as lter banks in subband coding. Discrete wavelet transform using Haar basis functions. Approximations (64 64, 128 128 and 256 256) that can be obtained from wavelet transform.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

18 / 20

Fast Wavelet Transform (FWT)

Analysis lter bank. Resulting decomposition. Synthesis lter bank.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

19 / 20

Wavelets in MATLAB (not mandatory)


1 2

Start MATLAB. Run wavemenu (part of wavelet toolbox). Select Wavelet-2D. Load an image from
/it/sw/matlab/current/... .../toolbox/wavelet/wavedemo/*.mat.

3 4

Experiment with different families, types, levels and results of different compression levels. Look at the families of wavelets by going to Display Wavelet Display.

M. Gavrilovic (Uppsala University)

L16 Wavelets and multiresolution

2008-05-22

20 / 20

Вам также может понравиться