Introduction to Wavelets Introduction to Wavelets Eric Arobone Eric Arobone
16

# Introduction to Wavelets - University of California, San Diego · Daughter wavelets All other wavelets are called daughter wavelets - defined in terms of the parent wavelets Notation

Apr 25, 2018

## Documents

truongtram
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript

Introduction to WaveletsIntroduction to Wavelets

Eric AroboneEric Arobone

What is a wavelet?What is a wavelet?

A basis function that is isolated with respect to - time or spatial location

- frequency or wavenumber

Each wavelet has a characteristic location and scale

Example wavelets (Haar)Example wavelets (Haar)

Parent waveletsFather wavelet () or scaling function

- Characterizes basic wavelet scale

- Covers entire domain of interest

Mother wavelet () or wavelet function

- Characterizes basic wavelet shape

- Covers entire domain of interest

Daughter waveletsDaughter wavelets

All other wavelets are called daughter wavelets

- defined in terms of the parent wavelets

Notation :

mu : directionality of wavelet functionsj : characteristic scale of waveleti's : horizontal and vertical shifts of wavelet functions

Directionality of wavelets???Directionality of wavelets???

What is a wavelet transform?What is a wavelet transform?

Representation of a function in real space as a linear combination of wavelet basis functions

Determining wavelet coefficientsDetermining wavelet coefficients

Wavelet coefficients are determined by an inner product relation (1D) :

In the discrete setting, the wavelet transform is computationally rather cheap : O(N)

- See references for implementation

Wavelet coefficientsWavelet coefficients

What makes a good wavelet?What makes a good wavelet?

Application specific, but in general...

Compact support

Orthogonality

Smoothness

Is there a contradiction here? Why?

Wavelet vs. Fourier transformWavelet vs. Fourier transformWavelet : spatial (time) and wavenumber (frequency) information

Fourier : wavenumber (frequency) information only

There is no free lunch

Wavelet : - not infinitely differentiable (smooth)- lose spectral accuracy when computing

derivatives- lose convolution theorem and other useful mathematical relationships

Why wavelets?Why wavelets?

Why perform a wavelet transform when thereare little to no simple mathematical operations in the wavelet basis?

Wavelet compressionWavelet compression

In many applications, wavelet transforms can be severely truncated (compressed) and retain useful information

Image compression- JPEG 2000

Signal compression

Video compression

Applications in fluid mechanicsApplications in fluid mechanics

Large Eddy Simulation (LES)- wavelet filtering can be used to extract energetic coherent structures from less energetic background flow

Compression of terabyte-sized datasets

Mixing layersMixing layersWavelet compression of vorticity fields has yielded great results (CVS)

Storing only 3.8% of wavelet coefficients, captures

- over 99% of turbulent kinetic energy

- over 83% of enstrophy

These results have motivated the use of wavelet PDE solvers for investigating turbulent flows

ReferencesReferencesBooks :

A First Course in Wavelets with Fourier Analysis, Boggess and Narcowich

Wavelets Make Easy, Nievergelt

Numerical Recipes in Fortran, Second Edition

Journals :

J. Fluid Mech. (2005), vol. 534, pp 39-66 (CVS)

Physics of Fluids 20, 045102 (2008)