Wavelet Toolbox™

Getting Started Guide

Michel Misiti
Yves Misiti
Georges Oppenheim
Jean-Michel Poggi

R2021a
How to Contact MathWorks

Latest news: www.mathworks.com

Sales and services: www.mathworks.com/sales_and_services

User community: www.mathworks.com/matlabcentral

Technical support: www.mathworks.com/support/contact_us

Phone: 508-647-7000

The MathWorks, Inc.

1 Apple Hill Drive
Natick, MA 01760-2098
Wavelet Toolbox™ Getting Started Guide
© COPYRIGHT 1997–2021 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used or copied
only under the terms of the license agreement. No part of this manual may be photocopied or reproduced in any form
without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by, for, or through
the federal government of the United States. By accepting delivery of the Program or Documentation, the government
hereby agrees that this software or documentation qualifies as commercial computer software or commercial computer
software documentation as such terms are used or defined in FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014.
Accordingly, the terms and conditions of this Agreement and only those rights specified in this Agreement, shall pertain
to and govern the use, modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government) and shall
supersede any conflicting contractual terms or conditions. If this License fails to meet the government's needs or is
inconsistent in any respect with federal procurement law, the government agrees to return the Program and
Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be
trademarks or registered trademarks of their respective holders.
Patents
MathWorks products are protected by one or more U.S. patents. Please see www.mathworks.com/patents for
more information.
Revision History
March 1997 First printing New for Version 1.0
September 2000 Second printing Revised for Version 2.0 (Release 12)
June 2001 Online only Revised for Version 2.1 (Release 12.1)
July 2002 Online only Revised for Version 2.2 (Release 13)
June 2004 Online only Revised for Version 3.0 (Release 14)
July 2004 Third printing Revised for Version 3.0
October 2004 Online only Revised for Version 3.0.1 (Release 14SP1)
March 2005 Online only Revised for Version 3.0.2 (Release 14SP2)
June 2005 Fourth printing Minor revision for Version 3.0.2
September 2005 Online only Minor revision for Version 3.0.3 (Release R14SP3)
March 2006 Online only Minor revision for Version 3.0.4 (Release 2006a)
September 2006 Online only Revised for Version 3.1 (Release 2006b)
March 2007 Online only Revised for Version 4.0 (Release 2007a)
September 2007 Online only Revised for Version 4.1 (Release 2007b)
October 2007 Fifth printing Revised for Version 4.1
March 2008 Online only Revised for Version 4.2 (Release 2008a)
October 2008 Online only Revised for Version 4.3 (Release 2008b)
March 2009 Online only Revised for Version 4.4 (Release 2009a)
September 2009 Online only Minor revision for Version 4.4.1 (Release 2009b)
March 2010 Online only Revised for Version 4.5 (Release 2010a)
September 2010 Online only Revised for Version 4.6 (Release 2010b)
April 2011 Online only Revised for Version 4.7 (Release 2011a)
September 2011 Online only Revised for Version 4.8 (Release 2011b)
March 2012 Online only Revised for Version 4.9 (Release 2012a)
September 2012 Online only Revised for Version 4.10 (Release 2012b)
March 2013 Online only Revised for Version 4.11 (Release 2013a)
September 2013 Online only Revised for Version 4.12 (Release 2013b)
March 2014 Online only Revised for Version 4.13 (Release 2014a)
October 2014 Online only Revised for Version 4.14 (Release 2014b)
March 2015 Online only Revised for Version 4.14.1 (Release 2015a)
September 2015 Online only Revised for Version 4.15 (Release 2015b)
March 2016 Online only Revised for Version 4.16 (Release 2016a)
September 2016 Online only Revised for Version 4.17 (Release 2016b)
March 2017 Online only Revised for Version 4.18 (Release 2017a)
September 2017 Online only Revised for Version 4.19 (Release 2017b)
March 2018 Online only Revised for Version 5.0 (Release 2018a)
September 2018 Online only Revised for Version 5.1 (Release 2018b)
March 2019 Online only Revised for Version 5.2 (Release 2019a)
September 2019 Online only Revised for Version 5.3 (Release 2019b)
March 2020 Online only Revised for Version 5.4 (Release 2020a)
September 2020 Online only Revised for Version 5.5 (Release 2020b)
March 2021 Online only Revised for Version 5.6 (Release 2021a)
 Contents

Getting Started with Wavelet Toolbox Software

1
Wavelet Toolbox Product Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2
Key Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2

Installing Wavelet Toolbox Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3

System Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3
Platform-Specific Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-3

What is a Wavelet? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5

Localize Discontinuity in Sine Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5

Choose a Wavelet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10

Time-Frequency Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-10
Multiresolution Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-13
Denoising . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-17
Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-17
General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-18

Extremal Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-21

Effect of Wavelet Support on Noisy Data . . . . . . . . . . . . . . . . . . . . . . . . . . 1-23

Comparing MODWT and MODWTMRA . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-27

Image Reconstruction with Biorthogonal Wavelets . . . . . . . . . . . . . . . . . 1-32

Wavelets and Vanishing Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-35

Continuous and Discrete Wavelet Transforms . . . . . . . . . . . . . . . . . . . . . 1-38

Guidelines for Continuous Wavelet Transform vs. Discrete Wavelet
Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-40

Nonstationary Gabor Frames and the Constant-Q Transform . . . . . . . . . 1-42

Decomposing the Time-Frequency Plane . . . . . . . . . . . . . . . . . . . . . . . . . 1-42
Constant-Q Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-43
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-44

From Fourier Analysis to Wavelet Analysis . . . . . . . . . . . . . . . . . . . . . . . . 1-46

Inner Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-46
Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-47
Short-Time Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-49

Continuous Wavelet Transform and Scale-Based Analysis . . . . . . . . . . . . 1-51

Definition of the Continuous Wavelet Transform . . . . . . . . . . . . . . . . . . . 1-51

v
 Scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-52
Shifting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-54
CWT as a Windowed Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-54

Continuous Wavelet Transform as a Bandpass Filter . . . . . . . . . . . . . . . . 1-56

CWT as a Filtering Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-56
DFT-Based Continuous Wavelet Transform . . . . . . . . . . . . . . . . . . . . . . . 1-58

Inverse Continuous Wavelet Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-59

Interpreting Continuous Wavelet Coefficients . . . . . . . . . . . . . . . . . . . . . 1-61

Critically-Sampled Discrete Wavelet Transform . . . . . . . . . . . . . . . . . . . . 1-73

One-Stage Filtering: Approximations and Details . . . . . . . . . . . . . . . . . . 1-73
Multiple-Level Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-75

Critically-Sampled Wavelet Reconstruction . . . . . . . . . . . . . . . . . . . . . . . 1-76

Reconstruction Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-76
Reconstructing Approximations and Details . . . . . . . . . . . . . . . . . . . . . . 1-77
Wavelets From Conjugate Mirror Filters . . . . . . . . . . . . . . . . . . . . . . . . . 1-78

Wavelet Synchrosqueezing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-80

What is Wavelet Synchrosqueezing? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-80
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-80
Required Component Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-81
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-82

Introduction to Wavelet Families . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-87

Haar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-87
Daubechies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-87
Biorthogonal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-88
Coiflets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-89
Symlets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-90
Morlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-90
Mexican Hat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-91
Meyer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-91
Other Real Wavelets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-91
Complex Wavelets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-92

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-93

Using Wavelets
2
Introduction to Wavelet Toolbox App and Functions . . . . . . . . . . . . . . . . . 2-2

Wavelets: Working with Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3

Understanding Images in the MATLAB Environment . . . . . . . . . . . . . . . . . 2-3
Indexed Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
Wavelet Decomposition of Indexed Images . . . . . . . . . . . . . . . . . . . . . . . . 2-4
RGB (Truecolor) Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
Wavelet Decomposition of Truecolor Images . . . . . . . . . . . . . . . . . . . . . . . 2-5

vi Contents
 Other Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
Image Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5

Density Estimation Using Wavelets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-8

1-D Estimation Using the Wavelet Analyzer App . . . . . . . . . . . . . . . . . . . . 2-8
Importing and Exporting Information from the Wavelet Analyzer App . . . 2-11

1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App . . . . 2-13

2-D Wavelet Coefficient Selection Using the Wavelet Analyzer App . . . . 2-20

1-D Extension Using the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . 2-25

2-D Extension Using the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . . 2-26

Image Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-27

Image Fusion Using the Command Line . . . . . . . . . . . . . . . . . . . . . . . . . 2-27
Image Fusion Using the Wavelet Analyzer App . . . . . . . . . . . . . . . . . . . . 2-29
Saving the Synthesized Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-32

1-D Fractional Brownian Motion Synthesis . . . . . . . . . . . . . . . . . . . . . . . . 2-33

New Wavelet for CWT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-35

Additional Wavelet GUIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-38

1-D Extension Using the Signal Extension Tool . . . . . . . . . . . . . . . . . . . . 2-38
2-D Extension Using the Image Extension Tool . . . . . . . . . . . . . . . . . . . . 2-41
Fractional Brownian Motion Synthesis Tool . . . . . . . . . . . . . . . . . . . . . . 2-42
New Wavelet for CWT Using the Pattern Adapted Admissible Wavelet
Design Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-45

Getting Started with Wavelet Analysis

3
Wavelet Families and Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2

Visualizing Wavelets, Wavelet Packets, and Wavelet Filters . . . . . . . . . . . 3-4

Continuous Wavelet Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7

Continuous Wavelet Transform and Inverse Continuous Wavelet

Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-9

Discrete Wavelet Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-13

1-D Wavelet Denoising . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-13
2-D Decimated Discrete Wavelet Analysis . . . . . . . . . . . . . . . . . . . . . . . . 3-15
Nondecimated Discrete Wavelet Analysis . . . . . . . . . . . . . . . . . . . . . . . . 3-17

Lifting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-20

Critically Sampled Wavelet Packet Analysis . . . . . . . . . . . . . . . . . . . . . . . 3-27

vii
 1

Getting Started with Wavelet Toolbox

Software

• “Wavelet Toolbox Product Description” on page 1-2

• “Installing Wavelet Toolbox Software” on page 1-3
• “What is a Wavelet?” on page 1-5
• “Choose a Wavelet” on page 1-10
• “Extremal Phase” on page 1-21
• “Effect of Wavelet Support on Noisy Data” on page 1-23
• “Comparing MODWT and MODWTMRA” on page 1-27
• “Image Reconstruction with Biorthogonal Wavelets” on page 1-32
• “Wavelets and Vanishing Moments” on page 1-35
• “Continuous and Discrete Wavelet Transforms” on page 1-38
• “Nonstationary Gabor Frames and the Constant-Q Transform” on page 1-42
• “From Fourier Analysis to Wavelet Analysis” on page 1-46
• “Continuous Wavelet Transform and Scale-Based Analysis” on page 1-51
• “Continuous Wavelet Transform as a Bandpass Filter” on page 1-56
• “Inverse Continuous Wavelet Transform” on page 1-59
• “Interpreting Continuous Wavelet Coefficients” on page 1-61
• “Critically-Sampled Discrete Wavelet Transform” on page 1-73
• “Critically-Sampled Wavelet Reconstruction” on page 1-76
• “Wavelet Synchrosqueezing” on page 1-80
• “Introduction to Wavelet Families” on page 1-87
• “References” on page 1-93
1 Getting Started with Wavelet Toolbox Software

Wavelet Toolbox Product Description

Analyze and synthesize signals and images using wavelets

Wavelet Toolbox provides functions and apps for analyzing and synthesizing signals and images. The
toolbox includes algorithms for continuous wavelet analysis, wavelet coherence, synchrosqueezing,
and data-adaptive time-frequency analysis. The toolbox also includes apps and functions for
decimated and nondecimated discrete wavelet analysis of signals and images, including wavelet
packets and dual-tree transforms.

Using continuous wavelet analysis, you can study the way spectral features evolve over time, identify
common time-varying patterns in two signals, and perform time-localized filtering. Using discrete
wavelet analysis, you can analyze signals and images at different resolutions to detect changepoints,
discontinuities, and other events not readily visible in raw data. You can compare signal statistics on
multiple scales, and perform fractal analysis of data to reveal hidden patterns.

With Wavelet Toolbox you can obtain a sparse representation of data, useful for denoising or
compressing the data while preserving important features. Many toolbox functions support C/C++
code generation for desktop prototyping and embedded system deployment.

Key Features
• Time-frequency analysis using continuous wavelet transform, wavelet coherence, constant-Q
transform, empirical mode decomposition, and Hilbert-Huang transform
• Wavelet Signal Denoiser app for denoising time-series data
• Decimated wavelet packet and wavelet transforms, including wavelet leaders for fractal analysis
• Nondecimated techniques, including dual-tree, stationary wavelet, maximal overlap discrete
wavelet, and wavelet packet transforms
• Signal, image denoising, and compression, including matching pursuit
• Lifting method for constructing custom wavelets

1-2
 Installing Wavelet Toolbox Software

Installing Wavelet Toolbox Software

To install this toolbox on your computer, see the appropriate platform-specific MATLAB® installation
guide. To determine if the Wavelet Toolbox software is already installed on your system, check for a
subfolder named wavelet within the main toolbox folder.

Wavelet Toolbox software can perform signal or image analysis. For indexed images or truecolor
images (represented by m-by-n-by-3 arrays of uint8), all wavelet functions use floating-point
operations. To avoid Out of Memory errors, be sure to allocate enough memory to process various
image sizes.

The memory can be real RAM or can be a combination of RAM and virtual memory. See your
operating system documentation for how to configure virtual memory.

System Recommendations
While not a requirement, we recommend you obtain Signal Processing Toolbox™ and Image
Processing Toolbox™ software to use in conjunction with the Wavelet Toolbox software. These
toolboxes provide complementary functionality that give you maximum flexibility in analyzing and
processing signals and images.

This manual makes no assumption that your computer is running any other MATLAB toolboxes.

Platform-Specific Details
Some details of the use of the Wavelet Toolbox software may depend on your hardware or operating
system.

Windows Fonts

We recommend you set your operating system to use “Small Fonts.” Set this option by clicking the
Display icon in your desktop's Control Panel (accessible through the Settings > Control Panel
submenu). Select the Configuration option, and then use the Font Size menu to change to Small
Fonts. You'll have to restart Windows® for this change to take effect.

Fonts for Non-Windows Platforms

We recommend you set your operating system to use standard default fonts.

However, for all platforms, if you prefer to use large fonts, some of the labels in the Wavelet Analyzer
app figures may be illegible when using the default display mode of the toolbox. To change the default
mode to accept large fonts, use the wtbxmngr function. (For more information, see either the
wtbxmngr help or its reference page.)

Mouse Compatibility

Wavelet Toolbox software was designed for three distinct types of mouse control.

Left Mouse Button Middle Mouse Button Right Mouse Button

Make selections. Activate Display cross-hairs to show position- Translate plots up and down,
controls. dependent information. and left and right.

1-3
1 Getting Started with Wavelet Toolbox Software

Left Mouse Button Middle Mouse Button Right Mouse Button

Note The functionality of the middle mouse button and the right mouse button can be inverted
depending on the platform.

1-4
 What is a Wavelet?

What is a Wavelet?
A wavelet is a waveform of effectively limited duration that has an average value of zero and nonzero
norm.

Many signals and images of interest exhibit piecewise smooth behavior punctuated by transients.
Speech signals are characterized by short bursts encoding consonants followed by steady-state
oscillations indicative of vowels. Natural images have edges. Financial time series exhibit transient
behavior, which characterize rapid upturns and downturns in economic conditions. Unlike the Fourier
basis, wavelet bases are adept at sparsely representing piecewise regular signals and images, which
include transient behavior.

Compare wavelets with sine waves, which are the basis of Fourier analysis. Sinusoids do not have
limited duration — they extend from minus to plus infinity. While sinusoids are smooth and
predictable, wavelets tend to be irregular and asymmetric.

Fourier analysis consists of breaking up a signal into sine waves of various frequencies. Similarly,
wavelet analysis is the breaking up of a signal into shifted and scaled versions of the original (or
mother) wavelet.

Just looking at pictures of wavelets and sine waves, you can see intuitively that signals with sharp
changes might be better analyzed with an irregular wavelet than with a smooth sinusoid.

It also makes sense that local features can be described better with wavelets that have local extent.
The following example illustrates this for a simple signal consisting of a sine wave with a
discontinuity.

Localize Discontinuity in Sine Wave

This example shows wavelet analysis can localize a discontinuity in a sine wave.

Create a 1-Hz sine wave sampled at 100 Hz. The duration of the sine wave is one second. The sine
wave has a discontinuity at t = 0 . 5 seconds.

t = linspace(0,1,100)';
x = sin(2*pi*t);
x1 = x-0.15;
y = zeros(size(x));
y(1:length(y)/2) = x(1:length(y)/2);
y(length(y)/2+1:end) = x1(length(y)/2+1:end);
stem(t,y,'markerfacecolor',[0 0 1]);
xlabel('Seconds');
ylabel('Amplitude');

1-5
1 Getting Started with Wavelet Toolbox Software

Obtain the nondecimated discrete wavelet transform of the sine wave using the 'sym2' wavelet and
plot the wavelet (detail) coefficients along with the original signal.

[swa,swd] = swt(y,1,'sym2');
subplot(211)
stem(t,y,'markerfacecolor',[0 0 1]);
title('Original Signal');
subplot(212)
stem(t,swd,'markerfacecolor',[0 0 1]);
title('Level 1 Wavelet Coefficients');

1-6
 What is a Wavelet?

Compare the Fourier coefficient magnitudes for the 1-Hz sine wave with and without the
discontinuity.

dftsig = fft([x y]);

dftsig = dftsig(1:length(y)/2+1,:);
df = 100/length(y);
freq = 0:df:50;
stem(freq,abs(dftsig));
xlabel('Hz'); ylabel('Magnitude');
legend('sine wave','sine wave with discontinuity');

1-7
1 Getting Started with Wavelet Toolbox Software

There is minimal difference in the magnitudes of the Fourier coefficients. Because the discrete
Fourier basis vectors have support over the entire time interval, the discrete Fourier transform does
not detect the discontinuity as efficiently as the wavelet transform.

Compare the level 1 wavelet coefficients for the sine wave with and without the discontinuity.

[swax,swdx] = swt(x,1,'sym2');
subplot(211)
stem(t,swd); title('Sine Wave with Discontinuity (Wavelet Coefficients)');
subplot(212)
stem(t,swdx); title('Sine Wave (Wavelet Coefficients)');

1-8
 What is a Wavelet?

The wavelet coefficients of the two signals demonstrate a significant difference. Wavelet analysis is
often capable of revealing characteristics of a signal or image that other analysis techniques miss,
like trends, breakdown points, discontinuities in higher derivatives, and self-similarity. Furthermore,
because wavelets provide a different view of data than those presented by Fourier techniques,
wavelet analysis can often significantly compress or denoise a signal without appreciable
degradation.

1-9
1 Getting Started with Wavelet Toolbox Software

Choose a Wavelet
There are two types of wavelet analysis: continuous and multiresolution. The type of wavelet analysis
best suited for your work depends on what you want to do with the data. This topic focuses on 1-D
data, but you can apply the same principles to 2-D data. To learn how to perform and interpret each
type of analysis, see “Practical Introduction to Continuous Wavelet Analysis” and “Practical
Introduction to Multiresolution Analysis”.

Time-Frequency Analysis
If your goal is to perform a detailed time-frequency analysis, choose the continuous wavelet
transform (CWT). In terms of implementation, scales are discretized more finely in the CWT than in
the discrete wavelet transform (DWT). For additional information, see “Continuous and Discrete
Wavelet Transforms” on page 1-38.

Instantaneous Frequency

The CWT is superior to the short-time Fourier transform (STFT) for signals in which the
instantaneous frequency grows rapidly. In the following figure, the instantaneous frequencies of the
hyperbolic chirp are plotted as dashed lines in the spectrogram and CWT-derived scalogram. For
additional information, see “Time-Frequency Analysis and Continuous Wavelet Transform”.

1-10
 Choose a Wavelet

Localizing Transients

The CWT is good at localizing transients in nonstationary signals. In the following figure, observe how
well the wavelet coefficients align with the abrupt changes that occur in the signal. For additional
information, see “Practical Introduction to Continuous Wavelet Analysis”.

1-11
1 Getting Started with Wavelet Toolbox Software

Supported Wavelets

To obtain the continuous wavelet transform of your data, use cwt and cwtfilterbank. Both
functions support the analytic wavelets listed in the following table. By default, cwt and
cwtfilterbank use the generalized Morse wavelet family. This family is defined by two parameters.
You can vary the parameters to recreate many commonly used wavelets. In the time-domain plots, the
red line and blue lines are the real and imaginary parts, respectively, of the wavelet. The contour
plots show the wavelet spread in time and frequency. For additional information, see “Morse
Wavelets” and “Generalized Morse and Analytic Morlet Wavelets”.

Wavelet Features Name Time Domain Time-Frequency

Domain
Generalized Morse Can vary two 'morse' (default)
Wavelet parameters to
change time and
frequency spread

1-12
 Choose a Wavelet

Wavelet Features Name Time Domain Time-Frequency

Domain
Analytic Morlet Equal variance in 'amor'
(Gabor) Wavelet time and frequency

Bump Wavelet Wider variance in 'bump'

time, narrower
variance in
frequency

All the wavelets in the table are analytic. Analytic wavelets are wavelets with one-sided spectra and
are complex valued in the time domain. These wavelets are a good choice for obtaining a time-
frequency analysis using the CWT. Because the wavelet coefficients are complex valued, the CWT
provides phase information. cwt and cwtfilterbank support analytic and anti-analytic wavelets.
For additional information, see “CWT-Based Time-Frequency Analysis”.

Multiresolution Analysis
In a multiresolution analysis (MRA), you approximate a signal at progressively coarser scales while
recording the differences between approximations at consecutive scales. You create the
approximations and the differences by taking the discrete wavelet transform (DWT) of the signal. The
DWT provides a sparse representation for many natural signals. Approximations are formed by
comparing the signal with scaled and translated copies of a scaling function. Differences between
consecutive scales, also known as details, are captured using scaled and translated copies of a
wavelet. On a log2 scale, the difference between consecutive scales is always 1. In the case of the
CWT, differences between consecutive scales are finer.

When generating the MRA, you can either subsample (decimate) the approximation by a factor of 2
every time you increase the scale or not. Each option offers advantages and disadvantages. If you
subsample, you end up with the same number of wavelet coefficients as the original signal. In the
decimated DWT, translations are integer multiples of scale. For the nondecimated DWT, translations
are integer shifts. A nondecimated DWT provides a redundant representation of the original data, but

1-13
1 Getting Started with Wavelet Toolbox Software

not as redundant as the CWT. Your application not only influences your choice of wavelet, but also
which version of the DWT to use.

Energy Preservation

If preserving energy in the analysis stage is important, you must use an orthogonal wavelet. An
orthogonal transform preserves energy. Consider using an orthogonal wavelet with compact support.
Keep in mind that except for the Haar wavelet, orthogonal wavelets with compact support are not
symmetric. The associated filters have nonlinear phase. This table lists supported orthogonal
wavelets. See wavemngr('read') for all wavelet family names.

Orthogonal Features Name See Also Representative

Wavelet
Coiflet Scaling function 'coifN' for N = coifwavf,
and wavelets have 1, 2, ..., 5 waveinfo
same number of
vanishing moments

Daubechies Nonlinear phase; 'dbN' for N = 1, dbaux, waveinfo,

energy 2, ..., 45 Extremal Phase
concentrated near Wavelet
the start of their Coefficients on
support page 1-21

Fejér-Korovkin Filters constructed 'fkN' for N = 4, fejerkorovkin,

to minimize the 6, 8, 14, 18, 22 waveinfo
difference between
a valid scaling
filter and the ideal
sinc lowpass filter;
are especially
useful in discrete
(decimated and
undecimated)
wavelet packet
transforms.

1-14
 Choose a Wavelet

Orthogonal Features Name See Also Representative

Wavelet
Haar Symmetric; special 'haar' ('db1') waveinfo
case of
Daubechies; useful
for edge detection

Symlet Least asymmetric; 'symN' for N = 2, symaux,

nearly linear phase 3, ..., 45 waveinfo, “Least
Asymmetric
Wavelet and
Phase”

Use waveinfo to learn more about individual wavelet families. For example, waveinfo('db').

Depending on how you address border distortions, the DWT might not conserve energy in the
analysis stage. For more information, see “Border Effects”. The maximal overlap discrete wavelet
transform modwt and maximal overlap discrete wavelet packet transform modwpt do conserve
energy. The wavelet packet decomposition dwpt does not conserve energy.

Feature Detection

If you want to find closely spaced features, choose wavelets with smaller support, such as haar, db2,
or sym2. The support of the wavelet should be small enough to separate the features of interest.
Wavelets with larger support tend to have difficulty detecting closely spaced features. Using wavelets
with large support can result in coefficients that do not distinguish individual features. In the
following figure, the upper plot shows a signal with spikes. The lower plot shows the first-level MRA
details of a maximal overlap DWT using the haar (thick blue lines) and db6 (thick red lines) wavelets.

1-15
1 Getting Started with Wavelet Toolbox Software

If your data has sparsely spaced transients, you can use wavelets with larger support.

Analysis of Variance

If your goal is to conduct an analysis of variance, the maximal overlap discrete wavelet transform
(MODWT) is suited for the task. The MODWT is a variation of the standard DWT.

• The MODWT conserves energy in the analysis stage.

• The MODWT partitions variance across scales. For examples, see “Wavelet Analysis of Financial
Data” and “Wavelet Changepoint Detection”.
• The MODWT requires an orthogonal wavelet, such as a Daubechies wavelet or Symlet.
• The MODWT is a shift-invariant transform. Shifting the input data shifts the wavelet coefficients
by an identical amount. The decimated DWT is not shift invariant. Shifting the input changes the
coefficients and can redistribute energy across scales.

See modwt, modwtmra, and modwtvar for more information. See also “Comparing MODWT and
MODWTMRA” on page 1-27.

Redundancy

Taking the decimated DWT, wavedec, of a signal using an orthonormal family of wavelets provides a
minimally redundant representation of the signal. There is no overlap in wavelets within and across
scales. The number of coefficients equals the number of signal samples. Minimally redundant
representations are a good choice for compression, when you want to remove features that are not
perceived.

The CWT of a signal provides a highly redundant representation of a signal. There is significant
overlap between wavelets within and across scales. Also, given the fine discretization of the scales,
the cost to compute the CWT and store the wavelet coefficients is significantly greater than the DWT.
The MODWT modwt is also a redundant transform but the redundancy factor is usually significantly
less than the CWT. Redundancy tends to reinforce signal characteristics and features you want to
examine, such as frequency breaks or other transient events.

If your work requires representing a signal with minimal redundancy, use wavedec. If your work
requires a redundant representation, use modwt or modwpt.

1-16
 Choose a Wavelet

Denoising
An orthogonal wavelet, such as a Symlet or Daubechies wavelet, is a good choice for denoising
signals. A biorthogonal wavelet can also be good for image processing. Biorthogonal wavelet filters
have linear phase which is very critical for image processing. Using a biorthogonal wavelet will not
introduce visual distortions in the image.

• An orthogonal transform does not color white noise. If white noise is provided as input to an
orthogonal transform, the output is white noise. Performing a DWT with a biorthogonal wavelet
colors white noise.
• An orthogonal transform preserves energy.

The sym4 wavelet is the default wavelet used in wdenoise and Wavelet Signal Denoiser app. The
bior4.4 biorthogonal wavelet is the default wavelet in wdenoise2.

Compression
If your work involves signal or image compression, consider using a biorthogonal wavelet. This table
lists the supported biorthogonal wavelets with compact support.

Biorthogonal Wavelet Features Name Representative

Biorthogonal Spline Compact support; 'biorNr.Nd' where
symmetric filters; linear Nr and Nd are the
phase numbers of vanishing
moments for the
reconstruction and
decomposition filters,
respectively; see
waveinfo('bior') for
supported values

1-17
1 Getting Started with Wavelet Toolbox Software

Biorthogonal Wavelet Features Name Representative

Reverse Biorthogonal Compact support; 'rbioNd.Nr' where
Spline symmetric filters; linear Nr and Nd are the
phase numbers of vanishing
moments for the
reconstruction and
decomposition filters,
respectively; see
waveinfo('rbio') for
supported values

Having two scaling function-wavelet pairs, one pair for analysis and another for synthesis, is useful
for compression.

• Biorthogonal wavelet filters are symmetric and have linear phase. (See “Least Asymmetric
Wavelet and Phase”.)
• The wavelets used for analysis can have many vanishing moments. A wavelet with N vanishing
moments is orthogonal to polynomials of degree N-1. Using a wavelet with many vanishing
moments results in fewer significant wavelet coefficients. Compression is improved.
• The dual wavelets used for synthesis can have better regularity. The reconstructed signal is
smoother.

Using an analysis filter with fewer vanishing moments than a synthesis filter can adversely affect
compression. For an example, see “Image Reconstruction with Biorthogonal Wavelets” on page 1-32.

When using biorthogonal wavelets, energy is not conserved at the analysis stage. See “Orthogonal
and Biorthogonal Filter Banks” for additional information.

General Considerations
Wavelets have properties that govern their behavior. Depending on what you want to do, some
properties can be more important.

Orthogonality

If a wavelet is orthogonal, the wavelet transform preserves energy. Except for the Haar wavelet, no
orthogonal wavelet with compact support is symmetric. The associated filter has nonlinear phase.

Vanishing Moments

A wavelet with N vanishing moments is orthogonal to polynomials of degree N−1. For an example,
see “Wavelets and Vanishing Moments” on page 1-35. The number of vanishing moments and the
oscillation of the wavelet have a loose relationship. As the number of vanishing moments grows, the
greater the wavelet oscillates.

1-18
 Choose a Wavelet

The number of vanishing moments also affects the support of a wavelet. Daubechies proved that a
wavelet with N vanishing moments must have a support of at least length 2N-1.

Names for many wavelets are derived from the number of vanishing moments. For example, db6 is
the Daubechies wavelet with six vanishing moments, and sym3 is the Symlet with three vanishing
moments. For coiflet wavelets, coif3 is the coiflet with six vanishing moments. For Fejér-Korovkin
wavelets, fk8 is the Fejér-Korovkin wavelet with a length 8 filter. Biorthogonal wavelet names are
derived from the number of vanishing moments the analysis wavelet and synthesis wavelet each have.
For instance, bior3.5 is the biorthogonal wavelet with three vanishing moments in the synthesis
wavelet and five vanishing moments in the analysis wavelet. To learn more, see waveinfo and
wavemngr.

If the number of vanishing moments N is equal to 1, 2, or 3, then dbN and symN are identical.

Regularity

Regularity is related to how many continuous derivatives a function has. Intuitively, regularity can be
considered a measure of smoothness. To detect an abrupt change in the data, a wavelet must be
sufficiently regular. For a wavelet to have N continuous derivatives, the wavelet must have at least N
+1 vanishing moments. See “Detecting Discontinuities and Breakdown Points” for an example. If your
data is relatively smooth with few transients, a more regular wavelet might be a better fit for your
work.

References
[1] Daubechies, Ingrid. Ten Lectures on Wavelets. Society for Industrial and Applied Mathematics,
1992.

See Also
Apps
Signal Multiresolution Analyzer | Wavelet Signal Denoiser

Functions
cwt | cwtfilterbank | dwtfilterbank | wavedec | wavedec2 | waveinfo | wavemngr |
wdenoise | wdenoise2

Related Examples
• “Practical Introduction to Multiresolution Analysis”
• “Practical Introduction to Continuous Wavelet Analysis”

1-19
1 Getting Started with Wavelet Toolbox Software

More About
• Understanding Wavelets, Part 1: What Are Wavelets
• Understanding Wavelets, Part 2: Types of Wavelet Transforms
• Understanding Wavelets, Part 3: An Example Application of the Discrete Wavelet Transform
• Understanding Wavelets, Part 4: An Example Application of the Continuous Wavelet Transform

1-20
 Extremal Phase

Extremal Phase
This example demonstrates that for a given support, the cumulative sum of the squared coefficients of
a scaling filter increase more rapidly for an extremal phase wavelet than other wavelets.

Generate the scaling filter coefficients for the db15 and sym15 wavelets. Both wavelets have support
of width 2 × 15 − 1 = 29.

[~,~,LoR_db,~] = wfilters('db15');
[~,~,LoR_sym,~] = wfilters('sym15');

Next, generate the scaling filter coefficients for the coif5 wavelet. This wavelet also has support of
width 6 × 5 − 1 = 29.

[~,~,LoR_coif,~] = wfilters('coif5');

Confirm the sum of the coefficients for all three wavelets equals 2.

sqrt(2)-sum(LoR_db)

ans = 2.2204e-16

sqrt(2)-sum(LoR_sym)

ans = 0

sqrt(2)-sum(LoR_coif)

ans = 2.2204e-16

Plot the cumulative sums of the squared coefficients. Note how rapidly the Daubechies sum
increases. This is because its energy is concentrated at small abscissas. Since the Daubechies wavelet
has extremal phase, the cumulative sum of its squared coefficients increases more rapidly than the
other two wavelets.

plot(cumsum(LoR_db.^2),'rx-')
hold on
plot(cumsum(LoR_sym.^2),'mo-')
plot(cumsum(LoR_coif.^2),'b*-')
legend('Daubechies','Symlet','Coiflet')
title('Cumulative Sum')

1-21
1 Getting Started with Wavelet Toolbox Software

See Also
dbaux | symaux

1-22
 Effect of Wavelet Support on Noisy Data

Effect of Wavelet Support on Noisy Data

In this example you demonstrate an instance of discontinuities in noisy data being represented more
sparsely using a Haar wavelet than when using a wavelet with larger support.

Create a noisy square wave with 512 samples. Plot the square wave.

n = 512;
t = 0:0.001:(n*0.001)-0.001;
yn = square(2*pi*10*t)+0.02*randn(size(t));
plot(yn)
grid on
title('Noisy Signal')

Obtain the maximal overlap discrete wavelet transform (MODWT) of the signal using the haar
wavelet. The haar wavelet has a support of length equal to 1

modhaar = modwt(yn,'haar');

Obtain the multiresolution analysis from the haar MODWT matrix and plot the first-level details.

mrahaar = modwtmra(modhaar,'haar');
stem(mrahaar(1,:),'Marker','none','ShowBaseLine','off');
grid on
title('First-Level MRA Details Using Haar Wavelet')

1-23
1 Getting Started with Wavelet Toolbox Software

Obtain the MODWT of the signal using the db6 wavelet. The db6 wavelet has a support of length
equal to 11.

moddb6 = modwt(yn,'db6');

Obtain the multiresolution analysis from the db6 MODWT matrix and plot the first-level details. The
discontinuities are represented less sparsely using the db6 wavelet than the haar wavelet.

mradb6 = modwtmra(moddb6,'db6');
stem(mradb6(1,:),'Marker','none','ShowBaseLine','off');
grid on
title('First-Level MRA Details Using db6 Wavelet')

1-24
 Effect of Wavelet Support on Noisy Data

Zoom in. Using the Haar wavelet results in fewer coefficients necessary to identify the change in the
signal.

ind = 40:110;
subplot(311)
plot(ind,yn(ind))
title('Close Up')
subplot(312)
stem(ind,mrahaar(1,ind),'Marker','none','ShowBaseLine','off','LineWidth',2)
title('Using Haar')
subplot(313)
stem(ind,mradb6(1,ind),'Marker','none','ShowBaseLine','off','LineWidth',2)
title('Using db6')

1-25
1 Getting Started with Wavelet Toolbox Software

See Also
modwt | modwtmra | waveinfo

1-26
 Comparing MODWT and MODWTMRA

Comparing MODWT and MODWTMRA

This example demonstrates the differences between the functions MODWT and MODWTMRA. The
MODWT partitions a signal's energy across detail coefficients and scaling coefficients. The
MODWTMRA projects a signal onto wavelet subspaces and a scaling subspace.

Choose the sym6 wavelet. Load and plot an electrocardiogram (ECG) signal. The sampling frequency
for the ECG signal is 180 hertz. The data are taken from Percival and Walden (2000), p.125 (data
originally provided by William Constantine and Per Reinhall, University of Washington).

load wecg
t = (0:numel(wecg)-1)/180;
wv = 'sym6';
plot(t,wecg)
grid on
title(['Signal Length = ',num2str(numel(wecg))])
xlabel('Time (s)')
ylabel('Amplitude')

Take the MODWT of the signal.

wtecg = modwt(wecg,wv);

The input data are samples of a function f (x) evaluated at N-many time points. The function can be
expressed as a linear combination of the scaling function ϕ(x) and wavelet ψ(x)at varying scales and

1-27
1 Getting Started with Wavelet Toolbox Software

N−1 J0 N−1
− J0 /2 − J0 − j/2 −j
translations: f (x) = ∑ ck 2 ϕ(2 x − k) + ∑ f j(x) where f j(x) = ∑ d j, k 2 ψ(2 x − k)
k=0 j=1 k=0
and J0 is the number of levels of wavelet decomposition. The first sum is the coarse scale
approximation of the signal, and the f j(x) are the details at successive scales. MODWT returns the N-
many coefficients {ck}and the ( J0 × N)-many detail coefficients {d j, k} of the expansion. Each row in
wtecg contains the coefficients at a different scale.

When taking the MODWT of a signal of length N, there are floor(log2(N))-many levels of
decomposition (by default). Detail coefficients are produced at each level. Scaling coefficients are
returned only for the final level. In this example, since N = 2048, J0 = floor(log2(2048)) = 11 and the
number of rows in wtecg is J0 + 1 = 11 + 1 = 12.

The MODWT partitions the energy across the various scales and scaling coefficients:
J0
2 2 2
|| X|| = ∑ | | W j | | + | | V J0 | | where X is the input data, W j are the detail coefficients at scale j,
j=1
and V J0 are the final-level scaling coefficients.

Compute the energy at each scale, and evaluate their sum.

energy_by_scales = sum(wtecg.^2,2);
Levels = {'D1';'D2';'D3';'D4';'D5';'D6';'D7';'D8';'D9';'D10';'D11';'A11'};
energy_table = table(Levels,energy_by_scales);
disp(energy_table)

Levels energy_by_scales
_______ ________________

{'D1' } 14.063
{'D2' } 20.612
{'D3' } 37.716
{'D4' } 25.123
{'D5' } 17.437
{'D6' } 8.9852
{'D7' } 1.2906
{'D8' } 4.7278
{'D9' } 12.205
{'D10'} 76.428
{'D11'} 76.268
{'A11'} 3.4192

energy_total = varfun(@sum,energy_table(:,2))

energy_total=table
sum_energy_by_scales
____________________

298.28

Confirm the MODWT is energy-preserving by computing the energy of the signal and comparing it
with the sum of the energies over all scales.

energy_ecg = sum(wecg.^2);
max(abs(energy_total.sum_energy_by_scales-energy_ecg))

1-28
 Comparing MODWT and MODWTMRA

ans = 7.4414e-10

Take the MODWTMRA of the signal.

mraecg = modwtmra(wtecg,wv);

MODWTMRA returns the projections of the function f (x) onto the various wavelet subspaces and final
N−1
− J0 /2 − J0
scaling space. That is, MODWTMRA returns ∑ ck 2 ϕ(2 x − k)and the J0-many {f j(x)}
k=0
evaluated at N-many time points. Each row in mraecg is a projection of f (x) onto a different
subspace. This means the original signal can be recovered by adding all the projections. This is not
true in the case of the MODWT. Adding the coefficients in wtecg will not recover the original signal.

Choose a time point, add the projections of f (x) evaluated at that time point and compare with the
original signal.

time_point = 1000;
abs(sum(mraecg(:,time_point))-wecg(time_point))

ans = 3.0849e-13

Confirm that, unlike MODWT, MODWTMRA is not an energy-preserving transform.

energy_ecg = sum(wecg.^2);
energy_mra_scales = sum(mraecg.^2,2);
energy_mra = sum(energy_mra_scales);
max(abs(energy_mra-energy_ecg))

ans = 115.7053

The MODWTMRA is a zero-phase filtering of the signal. Features will be time-aligned. Demonstrate
this by plotting the original signal and one of its projections. To better illustrate the alignment, zoom
in.

plot(t,wecg,'b')
hold on
plot(t,mraecg(4,:),'-')
hold off
grid on
xlim([4 8])
legend('Signal','Projection','Location','northwest')
xlabel('Time (s)')
ylabel('Amplitude')

1-29
1 Getting Started with Wavelet Toolbox Software

Make a similar plot using the MODWT coefficients at the same scale. Note that features will not be
time-aligned. The MODWT is not a zero-phase filtering of the input.

plot(t,wecg,'b')
hold on
plot(t,wtecg(4,:),'-')
hold off
grid on
xlim([4 8])
legend('Signal','Coefficients','Location','northwest')
xlabel('Time (s)')
ylabel('Amplitude')

1-30
 Comparing MODWT and MODWTMRA

References
[1] Percival, D. B., and A. T. Walden. Wavelet Methods for Time Series Analysis. Cambridge, UK:
Cambridge University Press, 2000.

See Also
modwt | modwtmra

1-31
1 Getting Started with Wavelet Toolbox Software

Image Reconstruction with Biorthogonal Wavelets

This example shows how applying the order biorthogonal wavelet filters can affect image
reconstruction.

Generate the analysis and synthesis filters for the bior3.5 wavelet. Load in and display an image.

[LoD,HiD,LoR,HiR] = wfilters('bior3.5');
load woman
imagesc(X)
colormap gray

The analysis filters, LoD and HiD, have 5 vanishing moments. The synthesis filters, LoR and HiR, have
3 vanishing moments. Do a five-level wavelet decomposition of the image using the analysis filters.

[c1,s1] = wavedec2(X,5,LoD,HiD);

Find the threshold that keeps only those wavelet coefficients with magnitudes in the top 10 percent.
Use the threshold to set the bottom 90 percent of coefficients to 0.

frac = 0.1;
c1sort = sort(abs(c1),'desc');
num = numel(c1);
thr = c1sort(floor(num*frac));
c1new = c1.*(abs(c1)>=thr);

1-32
 Image Reconstruction with Biorthogonal Wavelets

Reconstruct the image using the synthesis filters and the thresholded coefficients. Display the
reconstruction.

X1 = waverec2(c1new,s1,LoR,HiR);
figure
imagesc(X1)
colormap gray

Do a five-level wavelet decomposition of the image using the synthesis filters.

[c2,s2] = wavedec2(X,5,LoR,HiR);

Find the threshold that keeps only those wavelet coefficients with magnitudes in the top 10 percent.
Use the threshold to set the bottom 90 percent of coefficients to 0

frac = 0.1;
c2sort = sort(abs(c2),'desc');
num = numel(c2sort);
thr = c2sort(floor(num*frac));
c2new = c2.*(abs(c2)>=thr);

Reconstruct the image using the synthesis filters and the thresholded coefficients. Display the
reconstruction. Decomposing with a filter that has 3 vanishing moments and reconstructing with a
filter that has 5 vanishing moments results in poor reconstruction.

X2 = waverec2(c2new,s2,LoD,HiD);
figure

1-33
1 Getting Started with Wavelet Toolbox Software

imagesc(X2)
colormap gray

See Also
biorfilt | biorwavf | wavedec2 | waverec2 | wfilters

1-34
 Wavelets and Vanishing Moments

Wavelets and Vanishing Moments

This example shows how the number of vanishing moments can affect wavelet coefficients.

Create a signal defined over the interval 0 ≤ x ≤ 2. The signal is constant over the interval 0 ≤ x < 1
and quadratic over the interval 1 ≤ x ≤ 2. Plot the signal.
n = 1024;
x = linspace(0,2,n);
sig = zeros(1,n);
ind0 = (0<=x)&(x<1);
ind1 = (1<=x)&(x<=2);
sig(ind0) = 1;
sig(ind1) = x(ind1).^2;
plot(sig)
ylim([0 4])
grid on
title('Signal')

Compute a single-level wavelet decomposition of the signal using the db1 wavelet. This wavelet has
one vanishing moment. Plot the approximation coefficients and wavelet coefficients.
[a1,d1] = dwt(sig,'db1');
figure
subplot(2,1,1)
plot(a1)
ylim([0 6])

1-35
1 Getting Started with Wavelet Toolbox Software

grid on
title('Approximation Coefficients - db1')
subplot(2,1,2)
plot(d1)
ylim([-6e-3 0])
grid on
title('Wavelet Coefficients - db1')

The wavelet coefficients corresponding with the constant portion of the signal are approximately 0.
The magnitude of the wavelet coefficients corresponding with the quadratic portion of the signal are
increasing. Because the db1 wavelet has one vanishing moment, the wavelet is not orthogonal to the
quadratic portion of the signal.

Compute a single-level wavelet decomposition of the signal using the db3 wavelet. This wavelet has
three vanishing moments. Plot the approximation coefficients and wavelet coefficients.

[a2,d2] = dwt(sig,'db3');
figure
subplot(2,1,1)
plot(a2)
ylim([0 6])
grid on
title('Approximation Coefficients - db3')
subplot(2,1,2)
plot(d2)
grid on
title('Wavelet Coefficients - db3')

1-36
 Wavelets and Vanishing Moments

The wavelet coefficients corresponding with the constant portion of the signal are approximately 0.
The spike in the middle corresponds to where the constant and quadratic pieces of the signal meet.
The spike at the end is a boundary effect. The magnitude of the wavelet coefficients corresponding
with the quadratic portion of the signal are approximately 0. Because the db3 wavelet has three
vanishing moments, the wavelet is orthogonal to the quadratic part of the signal.

See Also
waveinfo

1-37
1 Getting Started with Wavelet Toolbox Software

Continuous and Discrete Wavelet Transforms

This topic describes the major differences between the continuous wavelet transform (CWT) and the
discrete wavelet transform (DWT) – both decimated and nondecimated versions. cwt is a discretized
version of the CWT so that it can be implemented in a computational environment. This discussion
focuses on the 1-D case, but is applicable to higher dimensions.

The cwt wavelet transform compares a signal with shifted and scaled (stretched or shrunk) copies of
1
a basic wavelet. If ψ(t) is a wavelet centered at t=0 with time support on [-T/2, T/2], then ψ( t −s u ) is
s
centered at t = u with time support [-sT/2+u, sT/2+u]. The cwt function uses L1 normalization so that
all frequency amplitudes are normalized to the same value. If 0<s<1, the wavelet is contracted
(shrunk) and if s>1, the wavelet is stretched. The mathematical term for this is dilation. See
“Continuous Wavelet Transform and Scale-Based Analysis” on page 1-51 for examples of how this
operation extracts features in the signal by matching it against dilated and translated wavelets.

The major difference between the CWT and discrete wavelet transforms, such as the dwt and modwt,
is how the scale parameter is discretized. The CWT discretizes scale more finely than the discrete
wavelet transform. In the CWT, you typically fix some base which is a fractional power of two, for
1/v
example, 2 where v is an integer greater than 1. The v parameter is often referred to as the
number of “voices per octave”. Different scales are obtained by raising this base scale to positive
j/v
integer powers, for example 2 j = 1, 2, 3, …. The translation parameter in the CWT is discretized to
integer values, denoted here by m. The resulting discretized wavelets for the CWT are

1
j/ν
ψ( n −j/vm ) .
2 2

The reason v is referred to as the number of voices per octave is because increasing the scale by an
v/v
octave (a doubling) requires v intermediate scales. Take for example 2 = 2 and then increase the
v/v 2v/v
numerator in the exponent until you reach 4, the next octave. You move from 2 = 2 to 2 = 4.
There are v intermediate steps. Common values for v are 10,12,14,16, and 32. The larger the value of
v, the finer the discretization of the scale parameter, s. However, this also increases the amount of
computation required because the CWT must be computed for every scale. The difference between
scales on a log2 scale is 1/v. See “CWT-Based Time-Frequency Analysis” and “Continuous Wavelet
Analysis of Modulated Signals” for examples of scale vectors with the CWT.

In the discrete wavelet transform, the scale parameter is always discretized to integer powers of 2, 2j,
j=1,2,3,..., so that the number of voices per octave is always 1. The difference between scales on a
log2 scale is always 1 for discrete wavelet transforms. Note that this is a much coarser sampling of
the scale parameter, s, than is the case with the CWT. Further, in the decimated (downsampled)
discrete wavelet transform (DWT), the translation parameter is always proportional to the scale. This
means that at scale, 2j, you always translate by 2jm where m is a nonnegative integer. In
nondecimated discrete wavelet transforms like modwt and swt, the scale parameter is restricted to
powers of two, but the translation parameter is an integer like in the CWT. The discretized wavelet
for the DWT takes the following form

1 j
ψ( 1j (n − 2 m)) .
j 2
2

The discretized wavelet for the nondecimated discrete wavelet transform, such as the MODWT, is

1-38
 Continuous and Discrete Wavelet Transforms

1
ψ( n −jm ) .
j 2
2

To summarize:

• The CWT and the discrete wavelet transforms differ in how they discretize the scale parameter.
The CWT typically uses exponential scales with a base smaller than 2, for example 21/12 . The
discrete wavelet transform always uses exponential scales with the base equal to 2. The scales in
the discrete wavelet transform are powers of 2. Keep in mind that the physical interpretation of
scales for both the CWT and discrete wavelet transforms requires the inclusion of the signal’s
sampling interval if it is not equal to one. For example, assume you are using the CWT and you set
1/12
your base to s0 = 2 . To attach physical significance to that scale, you must multiply by the
sampling interval Δt, so a scale vector covering approximately four octaves with the sampling
j
interval taken into account is s0 Δt j = 1, 2, ⋯48. Note that the sampling interval multiplies the
scales, it is not in the exponent. For discrete wavelet transforms the base scale is always 2.
• The decimated and nondecimated discrete wavelet transforms differ in how they discretize the
translation parameter. The decimated discrete wavelet transform (DWT), always translates by an
integer multiple of the scale, 2jm . The nondecimated discrete wavelet transform translates by
integer shifts.

These differences in how scale and translation are discretized result in advantages and disadvantages
for the two classes of wavelet transforms. These differences also determine use cases where one
wavelet transform is likely to provide superior results. Some important consequences of the
discretization of the scale and translation parameter are:

• The DWT provides a sparse representation for many natural signals. In other words, the important
features of many natural signals are captured by a subset of DWT coefficients that is typically
much smaller than the original signal. This “compresses” the signal. With the DWT, you always
end up with the same number of coefficients as the original signal, but many of the coefficients
may be close to zero in value. As a result, you can often throw away those coefficients and still
maintain a high-quality signal approximation. With the CWT, you go from N samples for an N-
length signal to a M-by-N matrix of coefficients with M equal to the number of scales. The CWT is
a highly redundant transform. There is significant overlap between wavelets at each scale and
between scales. The computational resources required to compute the CWT and store the
coefficients is much larger than the DWT. The nondecimated discrete wavelet transform is also
redundant but the redundancy factor is usually significantly less than the CWT, because the scale
parameter is not discretized so finely. For the nondecimated discrete wavelet transform, you go
from N samples to an L+1-by-N matrix of coefficients where L is the level of the transform.
• The strict discretization of scale and translation in the DWT ensures that the DWT is an
orthonormal transform (when using an orthogonal wavelet). There are many benefits of
orthonormal transforms in signal analysis. Many signal models consist of some deterministic
signal plus white Gaussian noise. An orthonormal transform takes this kind of signal and outputs
the transform applied to the signal plus white noise. In other words, an orthonormal transform
takes in white Gaussian noise and outputs white Gaussian noise. The noise is uncorrelated at the
input and output. This is important in many statistical signal processing settings. In the case of the
DWT, the signal of interest is typically captured by a few large-magnitude DWT coefficients, while
the noise results in many small DWT coefficients that you can throw away. If you have studied
linear algebra, you have no doubt learned many advantages to using orthonormal bases in the
analysis and representation of vectors. The wavelets in the DWT are like orthonormal vectors.
Neither the CWT nor the nondecimated discrete wavelet transform are orthonormal transforms.

1-39
1 Getting Started with Wavelet Toolbox Software

The wavelets in the CWT and nondecimated discrete wavelet transform are technically called
frames, they are linearly-dependent sets.
• The DWT is not shift-invariant. Because the DWT downsamples, a shift in the input signal does not
manifest itself as a simple equivalent shift in the DWT coefficients at all levels. A simple shift in a
signal can cause a significant realignment of signal energy in the DWT coefficients by scale. The
CWT and nondecimated discrete wavelet transform are shift-invariant. There are some
modifications of the DWT such as the dual-tree complex discrete wavelet transform that mitigate
the lack of shift invariance in the DWT, see “Critically Sampled and Oversampled Wavelet Filter
Banks” for some conceptual material on this topic and “Dual-Tree Complex Wavelet Transforms”
for an example.
• The discrete wavelet transforms are equivalent to discrete filter banks. Specifically, they are tree-
structured discrete filter banks where the signal is first filtered by a lowpass and a highpass filter
to yield lowpass and highpass subbands. Subsequently, the lowpass subband is iteratively filtered
by the same scheme to yield narrower octave-band lowpass and highpass subbands. In the DWT,
the filter outputs are downsampled at each successive stage. In the nondecimated discrete
wavelet transform, the outputs are not downsampled. The filters that define the discrete wavelet
transforms typically only have a small number of coefficients so the transform can be implemented
very efficiently. For both the DWT and nondecimated discrete wavelet transform, you do not
actually require an expression for the wavelet. The filters are sufficient. This is not the case with
the CWT. The most common implementation of the CWT requires you have the wavelet explicitly
defined. Even though the nondecimated discrete wavelet transform does not downsample the
signal, the filter bank implementation still allows for good computational performance, but not as
good as the DWT.
• The discrete wavelet transforms provide perfect reconstruction of the signal upon inversion. This
means that you can take the discrete wavelet transform of a signal and then use the coefficients to
synthesize an exact reproduction of the signal to within numerical precision. You can implement
an inverse CWT, but it is often the case that the reconstruction is not perfect. Reconstructing a
signal from the CWT coefficients is a much less stable numerical operation.
• The finer sampling of scales in the CWT typically results in a higher-fidelity signal analysis. You
can localize transients in your signal, or characterize oscillatory behavior better with the CWT
than with the discrete wavelet transforms.

For additional information on wavelet transforms and applications, see

• “From Fourier Analysis to Wavelet Analysis” on page 1-46

• “Continuous Wavelet Transform and Scale-Based Analysis” on page 1-51
• “Continuous Wavelet Transform as a Bandpass Filter” on page 1-56
• “Inverse Continuous Wavelet Transform” on page 1-59
• “Interpreting Continuous Wavelet Coefficients” on page 1-61
• “Critically-Sampled Discrete Wavelet Transform” on page 1-73
• “Wavelet Packets: Decomposing the Details”
• “Compare Time-Frequency Content in Signals with Wavelet Coherence”

Guidelines for Continuous Wavelet Transform vs. Discrete Wavelet

Transform
Based on the previous section, here are some basic guidelines for deciding on whether to use a
discrete or continuous wavelet transform.

1-40
 Continuous and Discrete Wavelet Transforms

• If your application is to obtain the sparsest possible signal representation for compression,
denoising, or signal transmission, use the DWT with wavedec.
• If your application requires an orthonormal transform, use the DWT with one of the orthogonal
wavelet filters. The orthogonal families in the Wavelet Toolbox are designated as type 1 wavelets
in the wavelet manager, wavemngr. Valid built-in orthogonal wavelet families are 'haar', 'dbN',
'fkN', 'coifN', or 'symN' where N is the number of vanishing moments for all families except
'fk'. For 'fk', N is the number of filter coefficients. See waveinfo for more detail.
• If your application requires a shift-invariant transform but you still need perfect reconstruction
and some measure of computational efficiency, try a nondecimated discrete wavelet transform like
modwt or a dual-tree transform like dualtree.
• If your primary goal is a detailed time-frequency (scale) analysis or precise localization of signal
transients, use cwt. For an example of time-frequency analysis with the CWT, see “CWT-Based
Time-Frequency Analysis”.
• For denoising a signal by thresholding wavelet coefficients, use the wdenoise function or the
Wavelet Signal Denoiser app. wdenoise and Wavelet Signal Denoiser provide default settings
that can be applied to your data, as well as a simple interface to a variety of denoising methods.
With the app, you can visualize and denoise signals, and compare results. For examples of
denoising a signal, see “Denoise A Signal Using Default Values” and “Denoise a Signal with the
Wavelet Signal Denoiser”. For denoising images, use wdenoise2. For an example, see “Denoising
Signals and Images”.
• If your application requires that you have a solid understanding of the statistical properties of the
wavelet coefficients, use a discrete wavelet transform. There is active work in understanding the
statistical properties of the CWT, but currently there are many more distributional results for the
discrete wavelet transforms. The success of the DWT in denoising is largely due to our
understanding of its statistical properties. For an example of estimation and hypothesis testing
using a nondecimated discrete wavelet transform see “Wavelet Analysis of Financial Data”.

See Also

Related Examples
• “Practical Introduction to Multiresolution Analysis”
• “Practical Introduction to Continuous Wavelet Analysis”

More About
• Understanding Wavelets, Part 1: What Are Wavelets
• Understanding Wavelets, Part 2: Types of Wavelet Transforms
• Understanding Wavelets, Part 3: An Example Application of the Discrete Wavelet Transform
• Understanding Wavelets, Part 4: An Example Application of the Continuous Wavelet Transform

1-41
1 Getting Started with Wavelet Toolbox Software

Nonstationary Gabor Frames and the Constant-Q Transform

In this section...
“Decomposing the Time-Frequency Plane” on page 1-42
“Constant-Q Transform” on page 1-43
“References” on page 1-44

Nonstationary Gabor frames enable you to implement time-adaptive or frequency-adaptive analysis of

signals. The functions cqt and icqt use nonstationary Gabor frames to obtain a constant-Q
(frequency-adaptive) transform (CQT) of a signal. A notable strength of nonstationary Gabor frames is
that they enable the construction of stable inverses, yielding perfect reconstruction.

The theory of nonstationary Gabor frames and efficient algorithms for their implementation are due
to Dörfler, Holighaus, Grill, and Velasco [1][2]. The algorithms in [1] and [2] implement a phase-
locked version of the CQT that does not preserve the same phases that would be obtained by naïve
convolution. In [3], Schörkhuber, Klapuri, Holighaus, and Dörfler develop efficient algorithms for the
CQT and inverse CQT that do mimic the coefficients obtained by naïve convolution. The Large Time-
Frequency Analysis Toolbox [4] provides an extensive set of algorithms for nonstationary Gabor
analysis and synthesis.

In standard Gabor analysis, a window of fixed size tiles the time-frequency plane. A nonstationary
Gabor frame is a collection of windowing functions of various sizes that are used to tile the time-
frequency plane. Wavelet analysis tiles the time-frequency plane in a similar manner. You have the
flexibility to change the sampling density in time or frequency. Nonstationary Gabor frames are useful
in areas such as audio signal processing, where fixed-sized time-frequency windows are not optimal.
Unlike the short-time Fourier transform, the windows used in the constant-Q transform have
adaptable bandwidth and sampling density. In frequency space, the windows are centered at
logarithmically spaced center frequencies.

Decomposing the Time-Frequency Plane

The Fourier transform of f(t) is the correlation of f(t) with ej ω t:

∫
∞
F(ω) = f (t)e− jωtdt .
−∞

Since ej ω t does not have compact support, the Fourier transform is not an ideal choice for studying
nonstationary signals. If the frequency content of a signal changes over time, the Fourier transform
does not capture what those changes are or when those changes occur. The partition of the time-
frequency plane shown here represents this Fourier transform behavior.

1-42
 Nonstationary Gabor Frames and the Constant-Q Transform

To perform a time-frequency analysis of a nonstationary signal, start with a real-valued even

windowing function, g(t), which is effectively nonzero over only a finite interval and has norm equal to
one. In addition, the Fourier transform of g(t) is centered at zero and is lowpass. Next, window f(t)
with translates of g(t). Then take the Fourier transform of the result

SF(u, ζ) = ∫ f (t)g(t − u)e − jζ tdt .

Correlating f(t) with the Gabor atoms, g(t − u)e jζt, is standard Gabor analysis. By varying u, you
consider only values of f(t) near time u. The support of g(t) determines the size of the neighborhood
near time u. The Fourier transform of gu, ζ (t) = g(t − u)eζt is the translation by ζ of the Fourier
transform of g(t) and is given by

g u, ζ (ω) = e−(ω − ζ)g (ω − ζ) .

The energy concentration of g u, ζ (ω) has variance σω and is centered at ζ. If the window,
gu, ζ (t) = g(t − u)eζt, shifts on a regular grid, the Fourier transform of the product of the shifted
window and f(t) is the short-time Fourier transform (STFT). The STFT tiling of the time-frequency
plane can be represented as a grid of boxes, each centered at (u, ζ):

The set of functions gu, ζ is known as a Gabor frame. The elements of this set are called Gabor
atoms. A frame is a set of functions, {hk(t)}, that satisfy the following condition: there exist constants
0 < A ≤ B < ∞ such that for any function f(t),
2 2 2
A f ≤ Σk f , hk ≤B f .

The energy concentration of g(t), in time, has variance σt. The energy concentration of g (ω), in
frequency, has variance σω. The energy concentration determines how well the window localizes the
signal in time and frequency. By the time-frequency uncertainty principle, there is a limit as to how
well you can simultaneously localize in both time and frequency domains, as indicated by

1
σtσω ≥ .
2

Narrowing the window in one domain results in poorer localization in the other domain. Gabor
showed that the area of the window is minimal when g(t) is Gaussian.

Constant-Q Transform
In the CQT, the bandwidth and sampling density in frequency are varied. The windows are
constructed and applied directly in the frequency domain. Different windows have different center

1-43
1 Getting Started with Wavelet Toolbox Software

frequencies and bandwidths, but the ratio of the center frequency to bandwidth remains constant.
Maintaining a constant ratio implies:

• Resolution in time improves at higher frequencies.

• Resolution in frequency improves at lower frequencies.

The time shifts for each window depend on the bandwidth, due to the uncertainty principle.

The CQT depends on:

• The window functions gk are real-valued, even functions. In the frequency domain, the Fourier
transform of gk is defined on the interval, [-Fs/2, Fs/2].
• The sampling rate, ζs.
• The number of bins per octave, b.
• The minimum and maximum frequencies, ζmin and ζmax.

Choose a minimum frequency ζmin and number of bins per octave b. Next, form a sequence of
geometrically spaced frequencies,
ζk = ζmin × 2k/b
for k = 0,...,K where K is an integer such that ζK is the largest frequency strictly less than the Nyquist
frequency ζs/2. The bandwidth at the kth frequency is set to Ωk = ζk+1-ζk-1. Given this sampling, the
ratio of the kth center frequency to the window bandwidth is independent of k:
Q = ζk/Δk = (21/b-2-1/b)-1.
To ensure perfect reconstruction, the DC component and Nyquist frequency are prepended and
appended, respectively, to the sequence.

W(ω) forms the window functions gk. W(ω) is a real-valued, even continuous function that is centered
at 0, positive in the interval [-½,½], and 0 elsewhere. W(ω) is translated to each center frequency ζk
then scaled. Evaluating a scaled and translated version of W(ω) yields the filter coefficients gk[m],
given by
gk[m] = W((m ζs/L - ζk)/Ωk)
for m = 0, …, L-1, where L is the signal length. By default, cqt uses the 'hann' window.

By the uncertainty principle, the size of the bandwidth constrains the value of the time shifts. To
satisfy the frame inequality, the shift akof gk must satisfy
ak ≤ ζk/Ωk.

As mentioned previously, the window is applied in the frequency domain. The filters, gk, centered at
ζk, are formed and applied to the Fourier transform of the signal. Taking the inverse transform
obtains the constant-Q coefficients.

References

[1] Holighaus, N., M. Dörfler, G.A. Velasco, and T. Grill. "A framework for invertible real-time
constant-Q transforms." IEEE Transactions on Audio, Speech, and Language Processing. Vol.
21, No. 4, 2013, pp. 775–785.

[2] Velasco, G. A., N. Holighaus, M. Dörfler, and T. Grill. "Constructing an invertible constant-Q
transform with nonstationary Gabor frames." In Proceedings of the 14th International
Conference on Digital Audio Effects (DAFx-11). Paris, France: 2011.

1-44
 Nonstationary Gabor Frames and the Constant-Q Transform

[3] Schörkhuber, C., A. Klapuri, N. Holighaus, and M. Dörfler. "A Matlab Toolbox for Efficient Perfect
Reconstruction Time-Frequency Transforms with Log-Frequency Resolution." Submitted to
the AES 53rd International Conference on Semantic Audio. London, UK: 2014.

[4] Průša, Z., P. L. Søndergaard, N. Holighaus, C. Wiesmeyr, and P. Balazs. The Large Time-Frequency
Analysis Toolbox 2.0. Sound, Music, and Motion, Lecture Notes in Computer Science 2014, pp
419-442. https://github.com/ltfat

See Also
cqt | icqt

More About
• “Short-Time Fourier Transform” on page 1-49

1-45
1 Getting Started with Wavelet Toolbox Software

From Fourier Analysis to Wavelet Analysis

In this section...
“Inner Products” on page 1-46
“Fourier Transform” on page 1-47
“Short-Time Fourier Transform” on page 1-49

Inner Products
Both the Fourier and wavelet transforms measure similarity between a signal and an analyzing
function. Both transforms use a mathematical tool called an inner product as this measure of
similarity. The two transforms differ in their choice of analyzing function. This results in the different
way the two transforms represent the signal and what kind of information can be extracted.

As a simple example of the inner product as a measure of similarity, consider the inner product of
vectors in the plane. The following MATLAB example calculates the inner product of three unit
vectors, u, v, w , in the plane:

3/2 1/ 2 0
, ,
1/2 1/ 2 1

u = [sqrt(3)/2 1/2];
v = [1/sqrt(2) 1/sqrt(2)];
w = [0 1];
% Three unit vectors in the plane
quiver([0 0 0],[0 0 0],[u(1) v(1) w(1)],[u(2) v(2) w(2)]);
axis([-1 1 0 1]);
text(-0.020,0.9371,'w');
text(0.6382,0.6623,'v');
text(0.7995,0.4751,'u');
% Compute inner products and print results
fprintf('The inner product of u and v is %1.2f\n', dot(u,v))
fprintf('The inner product of v and w is %1.2f\n', dot(w,v))
fprintf('The inner product of u and w is %1.2f\n', dot(u,w))

Looking at the figure, it is clear that u and v are most similar in their orientation, while u and w are
the most dissimilar.

1-46
 From Fourier Analysis to Wavelet Analysis

The inner products capture this geometric fact. Mathematically, the inner product of two vectors, u
and v is equal to the product of their norms and the cosine of the angle, θ, between them:

< u, v > = u v cos(θ)

For the special case when both u and v have unit norm, or unit energy, the inner product is equal to
cos(θ) and therefore lies between [-1,1]. In this case, you can interpret the inner product directly as a
correlation coefficient. If either u or v does not have unit norm, the inner product may exceed 1 in
absolute value. However, the inner product still depends on the cosine of the angle between the two
vectors making it interpretable as a kind of correlation. Note that the absolute value of the inner
product is largest when the angle between them is either 0 or π radians (0 or 180 degrees). This
occurs when one vector is a real-valued scalar multiple of the other.

While inner products in higher-dimensional spaces like those encountered in the Fourier and wavelet
transforms do not exhibit the same ease of geometric interpretation as the previous example, they
measure similarity in the same way. A significant part of the utility of these transforms is that they
essentially summarize the correlation between the signal and some basic functions with certain
physical properties, like frequency, scale, or position. By summarizing the signal in these constituent
parts, we are able to better understand the mechanisms that produced the signal.

Fourier Transform
Fourier analysis is used as a starting point to introduce the wavelet transforms, and as a benchmark
to demonstrate cases where wavelet analysis provides a more useful characterization of signals than
Fourier analysis.

Mathematically, the process of Fourier analysis is represented by the Fourier transform:

∫
∞
F(ω) = f (t)e− jωtdt .
−∞

which is the integral (sum) over all time of the signal f(t) multiplied by a complex exponential. Recall
that a complex exponential can be broken down into real and imaginary sinusoidal components. Note
that the Fourier transform maps a function of a single variable into another function of a single
variable.

1-47
1 Getting Started with Wavelet Toolbox Software

The integral defining the Fourier transform is an inner product. See “Inner Products” on page 1-46
for an example of how inner products measure of similarity between two signals. For each value of ω,
the integral (or sum) over all values of time produces a scalar, F(ω), that summarizes how similar the
two signals are. These complex-valued scalars are the Fourier coefficients. Conceptually, multiplying
each Fourier coefficient, F(ω), by a complex exponential (sinusoid) of frequency ω yields the
constituent sinusoidal components of the original signal. Graphically, the process looks like

Because e jωt is complex-valued, F(ω) is, in general, complex-valued. If the signal contains significant
oscillations at an angular frequency of ω0, the absolute value of F(ω0) will be large. By examining a
plot of F(ω) as a function of angular frequency, it is possible to determine what frequencies
contribute most to the variability of f(t).

To illustrate how the Fourier transform captures similarity between a signal and sinusoids of different
frequencies, the following MATLAB code analyzes a signal consisting of two sinusoids of 4 and 8
Hertz (Hz) corrupted by additive noise using the discrete Fourier transform.

rng(0,'twister');
Fs = 128;
t = linspace(0,1,128);
x = 2*cos(2*pi*4*t)+1.5*sin(2*pi*8*t)+randn(size(t));
xDFT = fft(x);
Freq = 0:64;
subplot(211);
plot(t,x); xlabel('Seconds'); ylabel('Amplitude');
subplot(212);
plot(Freq,abs(xDFT(1:length(xDFT)/2+1)))
set(gca,'xtick',[4:4:64]);
xlabel('Hz'); ylabel('Magnitude');

1-48
 From Fourier Analysis to Wavelet Analysis

Viewed as a time signal, it is difficult to determine what significant oscillations are present in the
data. However, looking at the absolute value of the Fourier transform coefficients as function of
frequency, the dominant oscillations at 4 and 8 Hz are easy to detect.

Short-Time Fourier Transform

The Fourier transform summarizes the similarity between a signal and a sinusoid with a single
complex number. The magnitude of the complex number captures the degree to which oscillations at
a particular frequency contribute to the signal's energy, while the argument of the complex number
captures phase information. Note that the Fourier coefficients have no time dependence. The Fourier
coefficients are obtained by integrating, or summing, over all time, so it is clear that this information
is lost. Consider the following two signals:

Both signals consist of a single sine wave with a frequency of 20 Hz. However, in the top signal, the
sine wave lasts the entire 1000 milliseconds. In the bottom plot, the sine wave starts at 250 and ends
at 750 milliseconds. The Fourier transform detects that the two signals have the same frequency
content, but has no way of capturing that the duration of the 20 Hz oscillation differs between the
two signals. Further, the Fourier transform has no mechanism for marking the beginning and end of
the intermittent sine wave.

In an effort to correct this deficiency, Dennis Gabor (1946) adapted the Fourier transform to analyze
only a small section of the signal at a time -- a technique called windowing the signal. Gabor's
adaptation is called the short-time Fourier transform (STFT). The technique works by choosing a time
function, or window, that is essentially nonzero only on a finite interval. As one example consider the
following Gaussian window function:

α −αt2
w(t) = e
π

The Gaussian function is centered around t=0 on an interval that depends on the value of α. Shifting
the Gaussian function by τ results in:

α −α(t − τ)2
w(t − τ) = e ,
π

1-49
1 Getting Started with Wavelet Toolbox Software

which centers the Gaussian window around τ. Multiplying a signal by w(t − τ) selects a portion of the
signal centered at τ. Taking the Fourier transform of these windowed segments for different values of
τ, produces the STFT. Mathematically, this is:

F(ω, τ) = ∫ f (t)w(t − τ)e

− jωtdt

The STFT maps a function of one variable into a function of two variables, ω and τ. This 2-D
representation of a 1-D signal means that there is redundancy in the STFT. The following figure
demonstrates how the STFT maps a signal into a time-frequency representation.

The STFT represents a sort of compromise between time- and frequency-based views of a signal. It
provides some information about both when and at what frequencies a signal event occurs. However,
you can only obtain this information with limited precision, and that precision is determined by the
size of the window.

While the STFT compromise between time and frequency information can be useful, the drawback is
that once you choose a particular size for the time window, that window is the same for all
frequencies. Many signals require a more flexible approach -- one where you can vary the window
size to determine more accurately either time or frequency.

Instead of plotting the STFT in three dimensions, the convention is to code F(ω, τ) as intensity on
some color map. Computing and displaying the STFT of the two 20-Hz sine waves of different
duration shown previously:

By using the STFT, you can see that the intermittent sine wave begins near 250 msec and ends
around 750 msec. Additionally, you can see that the signal's energy is concentrated around 20 Hz.

1-50
 Continuous Wavelet Transform and Scale-Based Analysis

Continuous Wavelet Transform and Scale-Based Analysis

In this section...
“Definition of the Continuous Wavelet Transform” on page 1-51
“Scale” on page 1-52
“Shifting” on page 1-54
“CWT as a Windowed Transform” on page 1-54

Definition of the Continuous Wavelet Transform

Like the Fourier transform, the continuous wavelet transform (CWT) uses inner products to measure
the similarity between a signal and an analyzing function. In the Fourier transform, the analyzing
functions are complex exponentials, e jωt. The resulting transform is a function of a single variable, ω.
In the short-time Fourier transform, the analyzing functions are windowed complex exponentials,
w(t)e jωt, and the result is a function of two variables. The STFT coefficients, F(ω, τ), represent the
match between the signal and a sinusoid with angular frequency ω in an interval of a specified length
centered at τ.

In the CWT, the analyzing function is a wavelet, ψ. The CWT compares the signal to shifted and
compressed or stretched versions of a wavelet. Stretching or compressing a function is collectively
referred to as dilation or scaling and corresponds to the physical notion of scale. By comparing the
signal to the wavelet at various scales and positions, you obtain a function of two variables. The 2-D
representation of a 1-D signal is redundant. If the wavelet is complex-valued, the CWT is a complex-
valued function of scale and position. If the signal is real-valued, the CWT is a real-valued function of
scale and position. For a scale parameter, a>0, and position, b, the CWT is:

∫
∞ 1 t−b
C(a, b; f (t), ψ(t)) = f (t) ψ* dt
−∞ a a

where * denotes the complex conjugate. Not only do the values of scale and position affect the CWT
coefficients, but the choice of wavelet also affects the values of the coefficients.

By continuously varying the values of the scale parameter, a, and the position parameter, b, you
obtain the cwt coefficients C(a,b). Note that for convenience, the dependence of the CWT coefficients
on the function and analyzing wavelet has been suppressed.

Multiplying each coefficient by the appropriately scaled and shifted wavelet yields the constituent
wavelets of the original signal.

There are many different admissible wavelets that can be used in the CWT. While it may seem
confusing that there are so many choices for the analyzing wavelet, it is actually a strength of wavelet
analysis. Depending on what signal features you are trying to detect, you are free to select a wavelet

1-51
1 Getting Started with Wavelet Toolbox Software

that facilitates your detection of that feature. For example, if you are trying to detect abrupt
discontinuities in your signal, you may choose one wavelet. On the other hand, if you are interested in
finding oscillations with smooth onsets and offsets, you are free to choose a wavelet that more closely
matches that behavior.

Scale
Like the concept of frequency, scale is another useful property of signals and images. For example,
you can analyze temperature data for changes on different scales. You can look at year-to-year or
decade-to-decade changes. Of course, you can examine finer (day-to-day), or coarser scale changes as
well. Some processes reveal interesting changes on long time, or spatial scales that are not evident
on small time or spatial scales. The opposite situation also happens. Some of our perceptual abilities
exhibit scale invariance. You recognize people you know regardless of whether you look at a large
portrait, or small photograph.

To go beyond colloquial descriptions such as “stretching” or “shrinking” we introduce the scale

factor, often denoted by the letter a. The scale factor is an inherently positive quantity, a>0. For
sinusoids, the effect of the scale factor is very easy to see.

In sin(at), the scale is the inverse of the radian frequency, a.

The scale factor works exactly the same with wavelets. The smaller the scale factor, the more
“compressed” the wavelet. Conversely, the larger the scale, the more stretched the wavelet. The
following figure illustrates this for wavelets at scales 1,2, and 4.

1-52
 Continuous Wavelet Transform and Scale-Based Analysis

This general inverse relationship between scale and frequency holds for signals in general.

Not only is a time-scale representation a different way to view data, but it also is a very natural way
to view data derived from a great number of natural phenomena.

Scale and Frequency

There is clearly a relationship between scale and frequency. Recall that longer scales correspond to
the most “stretched” wavelets. The more stretched the wavelet, the longer the portion of the signal
with which it is being compared, and therefore the coarser the signal features measured by the
wavelet coefficients.

To summarize, the general correspondence between scale and frequency is:

• Small scale a ⇒ Compressed wavelet ⇒ Rapidly changing details ⇒ High frequency ω.

• Long scale a ⇒ Stretched wavelet ⇒ Slowly changing, coarse features ⇒ Low frequency ω.

While there is a general relationship between scale and frequency, no precise relationship exists.
Users familiar with Fourier analysis often want to define a mapping between a wavelet at a given
scale with a specified sampling period to a frequency in hertz. You can only do this in a general sense.
Therefore, it is better to talk about the pseudo-frequency corresponding to a scale. The Wavelet

1-53
1 Getting Started with Wavelet Toolbox Software

Toolbox software provides two functions centfrq and scal2frq, which enable you to find these
approximate scale-frequency relationships for specified wavelets and scales.

The basic approach identifies the peak power in the Fourier transform of the wavelet as its center
frequency and divides that value by the product of the scale and the sampling interval. See scal2frq
for details. The following example shows the match between the estimated center frequency of the
db8 wavelet and a sinusoid of the same frequency.

The relationship between scale and frequency in the CWT is also explored in “Continuous Wavelet
Transform as a Bandpass Filter” on page 1-56.

Shifting
Shifting a wavelet simply means delaying (or advancing) its onset. Mathematically, delaying a
function f(t) by k is represented by f(t – k):

CWT as a Windowed Transform

In “Short-Time Fourier Transform” on page 1-49, the STFT is described as a windowing of the signal
to create a local frequency analysis. A shortcoming of the STFT approach is that the window size is
constant. There is a trade off in the choice of window size. A longer time window improves frequency
resolution while resulting in poorer time resolution because the Fourier transform loses all time
resolution over the duration of the window. Conversely, a shorter time window improves time
localization while resulting in poorer frequency resolution.

Wavelet analysis represents the next logical step: a windowing technique with variable-sized regions.
Wavelet analysis allows the use of long time intervals where you want more precise low-frequency
information, and shorter regions where you want high-frequency information.

1-54
 Continuous Wavelet Transform and Scale-Based Analysis

The following figure contrasts the different ways the STFT and wavelet analysis decompose the time-
frequency plane.

1-55
1 Getting Started with Wavelet Toolbox Software

Continuous Wavelet Transform as a Bandpass Filter

In this section...
“CWT as a Filtering Technique” on page 1-56
“DFT-Based Continuous Wavelet Transform” on page 1-58

CWT as a Filtering Technique

The continuous wavelet transform (CWT) computes the inner product of a signal, f (t), with translated
and dilated versions of an analyzing wavelet, ψ(t) . The definition of the CWT is:

∫
∞ 1 t−b
C(a, b; f (t), ψ(t)) = f (t) ψ* dt
−∞ a a

You can also interpret the CWT as a frequency-based filtering of the signal by rewriting the CWT as
an inverse Fourier transform.

∫
1 ∞
C(a, b; f (t), ψ(t)) = f (ω)ψ (aω)eiωbdω
2π −∞

where f (ω) and ψ (ω) are the Fourier transforms of the signal and the wavelet.

From the preceding equations, you can see that stretching a wavelet in time causes its support in the
frequency domain to shrink. In addition to shrinking the frequency support, the center frequency of
the wavelet shifts toward lower frequencies. The following figure demonstrates this effect for a
hypothetical wavelet and scale (dilation) factors of 1,2, and 4.

1-56
 Continuous Wavelet Transform as a Bandpass Filter

This depicts the CWT as a bandpass filtering of the input signal. CWT coefficients at lower scales
represent energy in the input signal at higher frequencies, while CWT coefficients at higher scales
represent energy in the input signal at lower frequencies. However, unlike Fourier bandpass filtering,
the width of the bandpass filter in the CWT is inversely proportional to scale. The width of the CWT
filters decreases with increasing scale. This follows from the uncertainty relationships between the
time and frequency support of a signal: the broader the support of a signal in time, the narrower its
support in frequency. The converse relationship also holds.

In the wavelet transform, the scale, or dilation operation is defined to preserve energy. To preserve
energy while shrinking the frequency support requires that the peak energy level increases. The
implementation of cwt in Wavelet Toolbox uses L1 normalization. The quality factor, or Q factor of a
filter is the ratio of its peak energy to bandwidth. Because shrinking or stretching the frequency
support of a wavelet results in commensurate increases or decreases in its peak energy, wavelets are
often referred to as constant-Q filters.

1-57
1 Getting Started with Wavelet Toolbox Software

DFT-Based Continuous Wavelet Transform

The equation in the preceding section defined the CWT as the inverse Fourier transform of a product
of Fourier transforms.
∞∧
C(a, b; f (t), ψ(t)) =
1
2π ∫
−∞
f (ω)ψ *(aω)e jωbdω

The time variable in the inverse Fourier transform is the translation parameter, b.

This suggests that you can compute the CWT with the inverse Fourier transform. Because there are
efficient algorithms for the computation of the discrete Fourier transform and its inverse, you can
often achieve considerable savings by using fft and ifft when possible.

To obtain a picture of the CWT in the Fourier domain, start with the definition of the wavelet
transform:

∫
1 ∞
< f (t), ψa, b(t) > = f (t)ψ*( t −a b )dt
a −∞

If you define:

1
ψa(t) = ψ*( − t/a)
a

you can rewrite the wavelet transform as

∫
∞
(f ∗ ψa)(b) = f (t)ψa(b − t)dt
−∞

which explicitly expresses the CWT as a convolution.

To implement the discretized version of the CWT, assume that the input sequence is a length N vector,
x[n]. The discrete version of the preceding convolution is:
N−1
Wa[b] = ∑ x[n] ψa[b − n]
n=0

To obtain the CWT, it appears you have to compute the convolution for each value of the shift
parameter, b, and repeat this process for each scale, a.

However, if the two sequences are circularly-extended (periodized to length N), you can express the
circular convolution as a product of discrete Fourier transforms. The CWT is the inverse Fourier
transform of the product
N−1
∧
1 2π ∧
N Δt k ∑
Wa(b) = X (2πk/NΔt)ψ * (a2πk/NΔt)e j2πkb/N
=0

where Δt is the sampling interval (period).

Expressing the CWT as an inverse Fourier transform enables you to use the computationally-efficient
fft and ifft algorithms to reduce the cost of computing convolutions.

The cwt function implements the CWT.

1-58
 Inverse Continuous Wavelet Transform

Inverse Continuous Wavelet Transform

The icwt function implements the inverse CWT. Using icwt requires that you obtain the CWT from
cwt.

Because the CWT is a redundant transform, there is not a unique way to define the inverse. The
inverse CWT implemented in Wavelet Toolbox uses the analytic Morse wavelet and L1 normalization.

The inverse CWT is classically presented in the double-integral form. Assume you have a wavelet ψ
with a Fourier transform that satisfies the admissibility condition:
∧ 2
ψ(ω)
∫
∞
Cψ = dω < ∞
−∞ ω

For wavelets satisfying the admissibility condition and finite-energy functions, f(t), you can define the
inverse CWT as:

f (t) =
1
Cψ ∫∫ < f (t), ψ
a b
a, b(t) > ψa, b(t) db
da
a2

1 t−b
where ψa, b(t) = ψ .
a a

For analyzing wavelets and functions satisfying the following conditions, a single integral formula for
the inverse CWT exists. These conditions are:

• The analyzed function, f(t), is real-valued and the analyzing wavelet has a real-valued Fourier
transform.
• The analyzed function, f(t), is real-valued and the Fourier transform of the analyzing wavelet has
support only on the set of nonnegative frequencies. This is referred to as an analytic wavelet. A
function whose Fourier transform only has support on the set of nonnegative frequencies must be
complex-valued.

The preceding conditions constrain the set of possible analyzing wavelets. Wavelets supported by cwt
are analytic. Because the toolbox only supports the analysis of real-valued functions, the real-valued
condition on the analyzed function is always satisfied.

To motivate the single integral formula, let ψ1 and ψ2 be two wavelets that satisfy the following two-
wavelet admissibility condition:
∧ ∧
ψ1*(ω) ψ2(ω)
∫ ω
dω < ∞

Define the constant:

∧ ∧
ψ1*(ω)ψ2(ω)
Cψ1, ψ2 = ∫ ω
dω

The above constant may be complex-valued. Let f(t) and g(t) be two finite energy functions. If the two-
wavelet admissibility condition is satisfied, the following equality holds:

Cψ1, ψ2 < f , g > = ∫∫ < f , ψ 1 > < g, ψ2 >* db da

a

1-59
1 Getting Started with Wavelet Toolbox Software

where < , > denotes the inner product, * denotes the complex conjugate, and the dependence of ψ1
and ψ2 on scale and position has been suppressed for convenience.

The key to the single integral formula for the inverse CWT is to recognize that the two-wavelet
admissibility condition can be satisfied even if one of the wavelets is not admissible. In other words, it
is not necessary that both ψ1 and ψ2 be separately admissible. You can also relax the requirements
further by allowing one of the functions and wavelets to be distributions. By first letting g(t) be the
Dirac delta function (a distribution) and also allowing ψ2 to be the Dirac delta function, you can
derive the single integral formula for the inverse CWT

∫
1 ∞ da
f (t) = 2 Re < f (t), ψ1(t) >
Cψ1, δ 0 a

where Re{ } denotes the real part.

The preceding equation demonstrates that you can reconstruct the signal by summing the scaled
CWT coefficients over all scales.

By summing the scaled CWT coefficients from select scales, you obtain an approximation to the
original signal. This is useful in situations where your phenomenon of interest is localized in scale.

icwt implements a discretized version of the above integral.

1-60
 Interpreting Continuous Wavelet Coefficients

Interpreting Continuous Wavelet Coefficients

Because the CWT is a redundant transform and the CWT coefficients depend on the wavelet, it can be
challenging to interpret the results.

To help you in interpreting CWT coefficients, it is best to start with a simple signal to analyze and an
analyzing wavelet with a simple structure.

A signal feature that wavelets are very good at detecting is a discontinuity, or singularity. Abrupt
transitions in signals result in wavelet coefficients with large absolute values.

For the signal create a shifted impulse. The impulse occurs at point 500.

x = zeros(1000,1);
x(500) = 1;

For the wavelet, pick the Haar wavelet.

[~,psi,xval] = wavefun('haar',10);
plot(xval,psi); axis([0 1 -1.5 1.5]);
title('Haar Wavelet');

To compute the CWT using the Haar wavelet at scales 1 to 128, enter:

CWTcoeffs = cwt(x,1:128,'haar');

CWTcoeffs is a 128-by-1000 matrix. Each row of the matrix contains the CWT coefficients for one
scale. There are 128 rows because the SCALES input to cwt is 1:128. The column dimension of the
matrix matches the length of the input signal.

Recall that the CWT of a 1D signal is a function of the scale and position parameters. To produce a
plot of the CWT coefficients, plot position along the x-axis, scale along the y-axis, and encode the
magnitude, or size of the CWT coefficients as color at each point in the x-y, or time-scale plane.

You can produce this plot using cwt with the optional input argument 'plot'.

cwt(x,1:128,'haar','plot');
colormap jet; colorbar;

1-61
1 Getting Started with Wavelet Toolbox Software

The preceding figure was modified with text labels to explicitly show which colors indicate large and
small CWT coefficients.

You can also plot the size of the CWT coefficients in 3D with

cwt(x,1:64,'haar','3Dplot'); colormap jet;

where the number of scales has been reduced to aid in visualization.

Examining the CWT of the shifted impulse signal, you can see that the set of large CWT coefficients is
concentrated in a narrow region in the time-scale plane at small scales centered around point 500. As
the scale increases, the set of large CWT coefficients becomes wider, but remains centered around
point 500. If you trace the border of this region, it resembles the following figure.

This region is referred to as the cone of influence of the point t=500 for the Haar wavelet. For a given
point, the cone of influence shows you which CWT coefficients are affected by the signal value at that
point.

To understand the cone of influence, assume that you have a wavelet supported on [-C, C]. Shifting
the wavelet by b and scaling by a results in a wavelet supported on [-Ca+b, Ca+b]. For the simple
case of a shifted impulse, δ(t − τ), the CWT coefficients are only nonzero in an interval around τ equal
to the support of the wavelet at each scale. You can see this by considering the formal expression of
the CWT of the shifted impulse.

∫
∞ 1 1
C(a, b; δ(t − τ), ψ(t)) = δ(t − τ) ψ*( t −a b )dt = ψ*( τ −a b )
−∞ a a

1-62
 Interpreting Continuous Wavelet Coefficients

For the impulse, the CWT coefficients are equal to the conjugated, time-reversed, and scaled wavelet
as a function of the shift parameter, b. You can see this by plotting the CWT coefficients for a select
few scales.

subplot(311)
plot(CWTcoeffs(10,:)); title('Scale 10');
subplot(312)
plot(CWTcoeffs(50,:)); title('Scale 50');
subplot(313)
plot(CWTcoeffs(90,:)); title('Scale 90');

The cone of influence depends on the wavelet. You can find and plot the cone of influence for a
specific wavelet with conofinf.

The next example features the superposition of two shifted impulses, δ(t − 300) + δ(t − 500). In this
case, use the Daubechies' extremal phase wavelet with four vanishing moments, db4. The following
figure shows the cone of influence for the points 300 and 500 using the db4 wavelet.

Look at point 400 for scale 20. At that scale, you can see that neither cone of influence overlaps the
point 400. Therefore, you can expect that the CWT coefficient will be zero at that point and scale. The
signal is only nonzero at two values, 300 and 500, and neither cone of influence for those values
includes the point 400 at scale 20. You can confirm this by entering:

x = zeros(1000,1);
x([300 500]) = 1;
CWTcoeffs = cwt(x,1:128,'db4');
plot(CWTcoeffs(20,:)); grid on;

1-63
1 Getting Started with Wavelet Toolbox Software

Next, look at the point 400 at scale 80. At scale 80, the cones of influence for both points 300 and 500
include the point 400. Even though the signal is zero at point 400, you obtain a nonzero CWT
coefficient at that scale. The CWT coefficient is nonzero because the support of the wavelet has
become sufficiently large at that scale to allow signal values 100 points above and below to affect the
CWT coefficient. You can confirm this by entering:

plot(CWTcoeffs(80,:));
grid on;

In the preceding example, the CWT coefficients became large in the vicinity of an abrupt change in
the signal. This ability to detect discontinuities is a strength of the wavelet transform. The preceding
example also demonstrated that the CWT coefficients localize the discontinuity best at small scales.
At small scales, the small support of the wavelet ensures that the singularity only affects a small set
of wavelet coefficients.

To demonstrate why the wavelet transform is so adept at detecting abrupt changes in the signal,
consider a shifted Heaviside, or unit step signal.

x = [zeros(500,1); ones(500,1)];
CWTcoeffs = cwt(x,1:64,'haar','plot'); colormap jet;

1-64
 Interpreting Continuous Wavelet Coefficients

Similar to the shifted impulse example, the abrupt transition in the shifted step function results in
large CWT coefficients at the discontinuity. The following figure illustrates why this occurs.

In the preceding figure, the red function is the shifted unit step function. The black functions labeled
A, B, and C depict Haar wavelets at the same scale but different positions. You can see that the CWT
coefficients around position A are zero. The signal is zero in that neighborhood and therefore the
wavelet transform is also zero because any wavelet integrates to zero.

Note the Haar wavelet centered around position B. The negative part of the Haar wavelet overlaps
with a region of the step function that is equal to 1. The CWT coefficients are negative because the
product of the Haar wavelet and the unit step is a negative constant. Integrating over that area yields
a negative number.

Note the Haar wavelet centered around position C. Here the CWT coefficients are zero. The step
function is equal to one. The product of the wavelet with the step function is equal to the wavelet.
Integrating any wavelet over its support is zero. This is the zero moment property of wavelets.

1-65
1 Getting Started with Wavelet Toolbox Software

At position B, the Haar wavelet has already shifted into the nonzero portion of the step function by
1/2 of its support. As soon as the support of the wavelet intersects with the unity portion of the step
function, the CWT coefficients are nonzero. In fact, the situation illustrated in the previous figure
coincides with the CWT coefficients achieving their largest absolute value. This is because the entire
negative deflection of the wavelet oscillation overlaps with the unity portion of the unit step while
none of the positive deflection of the wavelet does. Once the wavelet shifts to the point that the
positive deflection overlaps with the unit step, there will be some positive contribution to the integral.
The wavelet coefficients are still negative (the negative portion of the integral is larger in area), but
they are smaller in absolute value than those obtained at position B.

The following figure illustrates two other positions where the wavelet intersects the unity portion of
the unit step.

In the top figure, the wavelet has just begun to overlap with the unity portion of the unit step. In this
case, the CWT coefficients are negative, but not as large in absolute value as those obtained at
position B. In the bottom figure, the wavelet has shifted past position B and the positive deflection of
the wavelet begins to contribute to the integral. The CWT coefficients are still negative, but not as
large in absolute value as those obtained at position B.

1-66
 Interpreting Continuous Wavelet Coefficients

You can now visualize how the wavelet transform is able to detect discontinuities. You can also
visualize in this simple example exactly why the CWT coefficients are negative in the CWT of the
shifted unit step using the Haar wavelet. Note that this behavior differs for other wavelets.

x = [zeros(500,1); ones(500,1)];
CWTcoeffs = cwt(x,1:64,'haar','plot'); colormap jet;
% plot a few scales for visualization
subplot(311);
plot(CWTcoeffs(5,:)); title('Scale 5');
subplot(312);
plot(CWTcoeffs(10,:)); title('Scale 10');
subplot(313);
plot(CWTcoeffs(50,:)); title('Scale 50');

Next consider how the CWT represents smooth signals. Because sinusoidal oscillations are a common
phenomenon, this section examines how sinusoidal oscillations in the signal affect the CWT
coefficients. To begin, consider the sym4 wavelet at a specific scale superimposed on a sine wave.

Recall that the CWT coefficients are obtained by computing the product of the signal with the shifted
and scaled analyzing wavelet and integrating the result. The following figure shows the product of
the wavelet and the sinusoid from the preceding figure.

You can see that integrating over this product produces a positive CWT coefficient. That results
because the oscillation in the wavelet approximately matches a period of the sine wave. The wavelet
is in phase with the sine wave. The negative deflections of the wavelet approximately match the
negative deflections of the sine wave. The same is true of the positive deflections of both the wavelet
and sinusoid.

1-67
1 Getting Started with Wavelet Toolbox Software

The following figure shifts the wavelet 1/2 of the period of the sine wave.

Examine the product of the shifted wavelet and the sinusoid.

You can see that integrating over this product produces a negative CWT coefficient. That results
because the wavelet is 1/2 cycle out of phase with the sine wave. The negative deflections of the
wavelet approximately match the positive deflections of the sine wave. The positive deflections of the
wavelet approximately match the negative deflections of the sinusoid.

Finally, shift the wavelet approximately one quarter cycle of the sine wave.

The following figure shows the product of the shifted wavelet and the sinusoid.

1-68
 Interpreting Continuous Wavelet Coefficients

Integrating over this product produces a CWT coefficient much smaller in absolute value than either
of the two previous examples. That results because the negative deflection of the wavelet
approximately aligns with a positive deflection of the sine wave. Also, the main positive deflection of
the wavelet approximately aligns with a positive deflection of the sine wave. The resulting product
looks much more like a wavelet than the other two products. If it looked exactly like a wavelet, the
integral would be zero.

At scales where the oscillation in the wavelet occurs on either a much larger or smaller scale than the
period of the sine wave, you obtain CWT coefficients near zero. The following figure illustrates the
case where the wavelet oscillates on a much smaller scale than the sinusoid.

The product shown in the bottom pane closely resembles the analyzing wavelet. Integrating this
product results in a CWT coefficient near zero.

The following example constructs a 60-Hz sine wave and obtains the CWT using the sym8 wavelet.

t = linspace(0,1,1000);
x = cos(2*pi*60*t);
CWTcoeffs = cwt(x,1:64,'sym8','plot'); colormap jet;

1-69
1 Getting Started with Wavelet Toolbox Software

Note that the CWT coefficients are large in absolute value around scales 9 to 21. You can find the
pseudo-frequencies corresponding to these scales using the command:

freq = scal2frq(9:21,'sym8',1/1000);

Note that the CWT coefficients are large at scales near the frequency of the sine wave. You can
clearly see the sinusoidal pattern in the CWT coefficients at these scales with the following code.

surf(CWTcoeffs); colormap jet;

shading('interp'); view(-60,12);

The final example constructs a signal consisting of both abrupt transitions and smooth oscillations.
The signal is a 2-Hz sinusoid with two introduced discontinuities.

N = 1024;
t = linspace(0,1,1024);
x = 4*sin(4*pi*t);
x = x - sign(t - .3) - sign(.72 - t);
plot(t,x); xlabel('t'); ylabel('x');
grid on;

1-70
 Interpreting Continuous Wavelet Coefficients

Note the discontinuities near t=0.3 and t=0.7.

Obtain and plot the CWT using the sym4 wavelet.

CWTcoeffs = cwt(x,1:180,'sym4');
imagesc(t,1:180,abs(CWTcoeffs));
colormap jet; axis xy;
xlabel('t'); ylabel('Scales');

Note that the CWT detects both the abrupt transitions and oscillations in the signal. The abrupt
transitions affect the CWT coefficients at all scales and clearly separate themselves from smoother
signal features at small scales. On the other hand, the maxima and minima of the 2–Hz sinusoid are
evident in the CWT coefficients at large scales and not apparent at small scales.

The following general principles are important to keep in mind when interpreting CWT coefficients.

• Cone of influence— Depending on the scale, the CWT coefficient at a point can be affected by
signal values at points far removed. You have to take into account the support of the wavelet at
specific scales. Use conofinf to determine the cone of influence. Not all wavelets are equal in
their support. For example, the Haar wavelet has smaller support at all scales than the sym4
wavelet.
• Detecting abrupt transitions— Wavelets are very useful for detecting abrupt changes in a
signal. Abrupt changes in a signal produce relatively large wavelet coefficients (in absolute value)
centered around the discontinuity at all scales. Because of the support of the wavelet, the set of
CWT coefficients affected by the singularity increases with increasing scale. Recall this is the
definition of the cone of influence. The most precise localization of the discontinuity based on the
CWT coefficients is obtained at the smallest scales.

1-71
1 Getting Started with Wavelet Toolbox Software

• Detecting smooth signal features— Smooth signal features produce relatively large wavelet
coefficients at scales where the oscillation in the wavelet correlates best with the signal feature.
For sinusoidal oscillations, the CWT coefficients display an oscillatory pattern at scales where the
oscillation in the wavelet approximates the period of the sine wave.

The basic algorithm for the continuous wavelet transform (CWT) is:

1 Take a wavelet and compare it to a section at the start of the original signal.
2 Calculate a number, C, that represents how closely correlated the wavelet is with this section of
the signal. The larger the number C is in absolute value, the more the similarity. This follows from
the fact the CWT coefficients are calculated with an inner product. See “Inner Products” on page
1-46 for more information on how inner products measure similarity. If the signal energy and the
wavelet energy are equal to one, C may be interpreted as a correlation coefficient. Note that, in
general, the signal energy does not equal one and the CWT coefficients are not directly
interpretable as correlation coefficients.

As described in “Continuous and Discrete Wavelet Transforms” on page 1-38, the CWT
coefficients explicitly depend on the analyzing wavelet. Therefore, the CWT coefficients are
different when you compute the CWT for the same signal using different wavelets.

3 Shift the wavelet to the right and repeat steps 1 and 2 until you've covered the whole signal.

4 Scale (stretch) the wavelet and repeat steps 1 through 3.

5 Repeat steps 1 through 4 for all scales.

1-72
 Critically-Sampled Discrete Wavelet Transform

Critically-Sampled Discrete Wavelet Transform

Calculating wavelet coefficients at every possible scale is a fair amount of work, and it generates an
awful lot of data. What if we choose only a subset of scales and positions at which to make our
calculations?

It turns out, rather remarkably, that if we choose scales and positions based on powers of two — so-
called dyadic scales and positions — then our analysis will be much more efficient and just as
accurate. We obtain such an analysis from the discrete wavelet transform (DWT). For more
information on DWT, see “Algorithms” in the Wavelet Toolbox User's Guide.

An efficient way to implement this scheme using filters was developed in 1988 by Mallat (see [Mal89]
in “References” on page 1-93). The Mallat algorithm is in fact a classical scheme known in the signal
processing community as a two-channel subband coder (see page 1 of the book Wavelets and Filter
Banks, by Strang and Nguyen [StrN96]).

This very practical filtering algorithm yields a fast wavelet transform — a box into which a signal
passes, and out of which wavelet coefficients quickly emerge. Let's examine this in more depth.

One-Stage Filtering: Approximations and Details

For many signals, the low-frequency content is the most important part. It is what gives the signal its
identity. The high-frequency content, on the other hand, imparts flavor or nuance. Consider the
human voice. If you remove the high-frequency components, the voice sounds different, but you can
still tell what's being said. However, if you remove enough of the low-frequency components, you hear
gibberish.

In wavelet analysis, we often speak of approximations and details. The approximations are the high-
scale, low-frequency components of the signal. The details are the low-scale, high-frequency
components.

The filtering process, at its most basic level, looks like this.

The original signal, S, passes through two complementary filters and emerges as two signals.

Unfortunately, if we actually perform this operation on a real digital signal, we wind up with twice as
much data as we started with. Suppose, for instance, that the original signal S consists of 1000
samples of data. Then the resulting signals will each have 1000 samples, for a total of 2000.

These signals A and D are interesting, but we get 2000 values instead of the 1000 we had. There
exists a more subtle way to perform the decomposition using wavelets. By looking carefully at the
computation, we may keep only one point out of two in each of the two 2000-length samples to get

1-73
1 Getting Started with Wavelet Toolbox Software

the complete information. This is the notion of downsampling. We produce two sequences called cA
and cD.

The process on the right, which includes downsampling, produces DWT coefficients.

To gain a better appreciation of this process, let's perform a one-stage discrete wavelet transform of a
signal. Our signal will be a pure sinusoid with high-frequency noise added to it.

Here is our schematic diagram with real signals inserted into it.

The MATLAB code needed to generate s, cD, and cA is

s
= sin(20.*linspace(0,pi,1000)) + 0.5.*rand(1,1000);
[cA,cD] = dwt(s,'db2');

where db2 is the name of the wavelet we want to use for the analysis.

Notice that the detail coefficients cD are small and consist mainly of a high-frequency noise, while the
approximation coefficients cA contain much less noise than does the original signal.

[length(cA) length(cD)]

ans =
501 501

You may observe that the actual lengths of the detail and approximation coefficient vectors are
slightly more than half the length of the original signal. This has to do with the filtering process,
which is implemented by convolving the signal with a filter. The convolution “smears” the signal,
introducing several extra samples into the result.

1-74
 Critically-Sampled Discrete Wavelet Transform

Multiple-Level Decomposition
The decomposition process can be iterated, with successive approximations being decomposed in
turn, so that one signal is broken down into many lower resolution components. This is called the
wavelet decomposition tree.

Looking at a signal's wavelet decomposition tree can yield valuable information.

Number of Levels

Since the analysis process is iterative, in theory it can be continued indefinitely. In reality, the
decomposition can proceed only until the individual details consist of a single sample or pixel. In
practice, you'll select a suitable number of levels based on the nature of the signal, or on a suitable
criterion such as entropy (see “Choosing the Optimal Decomposition” in the Wavelet Toolbox User's
Guide).

1-75
1 Getting Started with Wavelet Toolbox Software

Critically-Sampled Wavelet Reconstruction

We've learned how the discrete wavelet transform can be used to analyze, or decompose, signals and
images. This process is called decomposition or analysis. The other half of the story is how those
components can be assembled back into the original signal without loss of information. This process
is called reconstruction, or synthesis. The mathematical manipulation that effects synthesis is called
the inverse discrete wavelet transform (IDWT).

To synthesize a signal using Wavelet Toolbox software, we reconstruct it from the wavelet
coefficients.

Where wavelet analysis involves filtering and downsampling, the wavelet reconstruction process
consists of upsampling and filtering. Upsampling is the process of lengthening a signal component by
inserting zeros between samples.

The toolbox includes commands, like idwt and waverec, that perform single-level or multilevel
reconstruction, respectively, on the components of 1-D signals. These commands have their 2-D and
3-D analogs, idwt2, waverec2, idwt3, and waverec3.

Reconstruction Filters
The filtering part of the reconstruction process also bears some discussion, because it is the choice of
filters that is crucial in achieving perfect reconstruction of the original signal.

The downsampling of the signal components performed during the decomposition phase introduces a
distortion called aliasing. It turns out that by carefully choosing filters for the decomposition and
reconstruction phases that are closely related (but not identical), we can “cancel out” the effects of
aliasing.

A technical discussion of how to design these filters is available on page 347 of the book Wavelets and
Filter Banks, by Strang and Nguyen. The low- and high-pass decomposition filters (L and H), together
with their associated reconstruction filters (L' and H'), form a system of what is called quadrature
mirror filters:

1-76
 Critically-Sampled Wavelet Reconstruction

Reconstructing Approximations and Details

We have seen that it is possible to reconstruct our original signal from the coefficients of the
approximations and details.

It is also possible to reconstruct the approximations and details themselves from their coefficient
vectors. As an example, let's consider how we would reconstruct the first-level approximation A1 from
the coefficient vector cA1.

We pass the coefficient vector cA1 through the same process we used to reconstruct the original
signal. However, instead of combining it with the level-one detail cD1, we feed in a vector of zeros in
place of the detail coefficients vector:

The process yields a reconstructed approximation A1, which has the same length as the original
signal S and which is a real approximation of it.

Similarly, we can reconstruct the first-level detail D1, using the analogous process:

1-77
1 Getting Started with Wavelet Toolbox Software

The reconstructed details and approximations are true constituents of the original signal. In fact, we
find when we combine them that

A1 + D1 = S.

Note that the coefficient vectors cA1 and cD1 — because they were produced by downsampling and
are only half the length of the original signal — cannot directly be combined to reproduce the signal.
It is necessary to reconstruct the approximations and details before combining them.

Extending this technique to the components of a multilevel analysis, we find that similar relationships
hold for all the reconstructed signal constituents. That is, there are several ways to reassemble the
original signal:

Wavelets From Conjugate Mirror Filters

In the section “Reconstruction Filters” on page 1-76, we spoke of the importance of choosing the
right filters. In fact, the choice of filters not only determines whether perfect reconstruction is
possible, it also determines the shape of the wavelet we use to perform the analysis.

To construct a wavelet of some practical utility, you seldom start by drawing a waveform. Instead, it
usually makes more sense to design the appropriate quadrature mirror filters, and then use them to
create the waveform. Let's see how this is done by focusing on an example.

Consider the low-pass reconstruction filter (L') for the db2 wavelet.

The filter coefficients can be obtained from the dbaux function. By reversing the order of the scaling
filter vector and multiplying every even element (indexing from 1) by (-1), you obtain the high-pass
filter.

Repeatedly upsampling by two and convolving the output with the scaling filter produces the
Daubechies' extremal phase wavelet.

L = dbaux(2);
H = wrev(L).*[1 -1 1 -1];
HU = dyadup(H,0);
HU = conv(HU,L);
plot(HU); title('1st Iteration');
H1 = conv(dyadup(HU,0),L);
H2 = conv(dyadup(H1,0),L);
H3 = conv(dyadup(H2,0),L);
H4 = conv(dyadup(H3,0),L);
figure;
for k =1:4
subplot(2,2,k);
eval(['plot(H' num2str(k) ')']);

1-78
 Critically-Sampled Wavelet Reconstruction

axis tight;
end

The curve begins to look progressively more like the db2 wavelet. This means that the wavelet's
shape is determined entirely by the coefficients of the reconstruction filters.

This relationship has profound implications. It means that you cannot choose just any shape, call it a
wavelet, and perform an analysis. At least, you can't choose an arbitrary wavelet waveform if you
want to be able to reconstruct the original signal accurately. You are compelled to choose a shape
determined by quadrature mirror decomposition filters.

Scaling Function

We've seen the interrelation of wavelets and quadrature mirror filters. The wavelet function ψ is
determined by the high-pass filter, which also produces the details of the wavelet decomposition.

There is an additional function associated with some, but not all, wavelets. This is the so-called
scaling function, ϕ. The scaling function is very similar to the wavelet function. It is determined by
the low-pass quadrature mirror filters, and thus is associated with the approximations of the wavelet
decomposition.

In the same way that iteratively upsampling and convolving the high-pass filter produces a shape
approximating the wavelet function, iteratively upsampling and convolving the low-pass filter
produces a shape approximating the scaling function.

1-79
1 Getting Started with Wavelet Toolbox Software

Wavelet Synchrosqueezing
What is Wavelet Synchrosqueezing?
The wavelet synchrosqueezed transform is a time-frequency analysis method that is useful for
analyzing multicomponent signals with oscillating modes. Examples of signals with oscillating modes
include speech waveforms, machine vibrations, and physiologic signals. Many of these real-world
signals with oscillating modes can be written as a sum of amplitude-modulated and frequency-
modulated components. A general expression for these types of signals with summed components is
K
∑ Ak(t)cos(2πϕk(t)),
k=1

where Ak(t) is the slowly varying amplitude and ϕk(t) is the instantaneous phase. A truncated Fourier
series, where the amplitude and frequency do not vary with time, is a special case of these signals.

The wavelet transform and other linear time-frequency analysis methods decompose these signals
into their components by correlating the signal with a dictionary of time-frequency atoms [1]. The
wavelet transform uses translated and scaled versions of a mother wavelet as its time-frequency
atom. Some time-frequency spreading is associated with all of these time-frequency atoms, which
affects the sharpness of the signal analysis.

The wavelet synchrosqueezed transform is a time-frequency method that reassigns the signal energy
in frequency. This reassignment compensates for the spreading effects caused by the mother wavelet.
Unlike other time-frequency reassignment methods, synchrosqueezing reassigns the energy only in
the frequency direction, which preserves the time resolution of the signal. By preserving the time, the
inverse synchrosqueezing algorithm can reconstruct an accurate representation of the original signal.
To use synchrosqueezing, each term in the summed components signal expression must be an
intrinsic mode type (IMT) function. For details on the criteria that constitute IMTs, see [2].

Algorithm
The synchrosqueezing algorithm uses these steps.
1 Obtain the CWT of the input signal. For use with synchrosqueezing, the CWT must use an
analytic wavelet to capture instantaneous frequency information.
2 Extract the instantaneous frequencies from the CWT output, W f , using a phase transform, ωf .
This phase transform is proportional to the first derivative of the CWT with respect to the
translation, u. In this definition of the phase transform, s is the scale.

∂tW f (s, u)
ωf (s, u) = .
2πiW f (s, u)

fχ
The scales are defined as s = , where f χ is the peak frequency and f is the frequency.
f
i2πf 0t
To extract the instantaneous frequency, consider a simple sine wave, e .
a Obtain the wavelet transform,
i2πf 0t i2πf 0u
Wf e =e ,

1-80
 Wavelet Synchrosqueezing

where χ sf 0 is the Fourier transform of the wavelet at sf0.

b Take the partial derivative of the previous equation with respect to the translation, u:

∂ i2πf 0t i2πf 0u
W e = i2πf 0 χ f χ e
∂u f
c Divide the partial derivative by the wavelet transform and i2π to obtain the instantaneous
frequency, f0.
3 “Squeeze” the CWT over regions where the phase transform is constant. The resulting
instantaneous frequency value is reassigned to a single value at the centroid of the CWT time-
frequency region. This reassignment results in sharpened output from the synchrosqueezed
transform when compared to the CWT.

As described, synchrosqueezing uses the continuous wavelet transform (CWT) and its first derivative
with respect to translation. The CWT is invertible and since the synchrosqueezed transform inherits
the CWT invertibility property, the signal can be reconstructed.

Required Component Separation

With synchrosqueezing the signal components must be IMTs that are well separated in the time-
frequency plane. If this requirement is met, you can track the trajectory of the instantaneous
frequencies along a curve. The curves show the location of the maximum energy as it varies over time
for each signal mode. See wsstridge for a description of the trajectory curves algorithm.

This inequality defines the required separation criteria:

1
ϕk′ (t) − ϕk′ − 1(t) ≥ ϕ′ (t) + ϕk′ − 1(t)
4 k

where ϕ′ is the instantaneous frequency and d is a positive separation constant [2]. To determine this
required separation, suppose a bump wavelet, x, has a Fourier transform with support in the range
5 5 1 5 1
εx − Δ, εx + Δ . Because the bump wavelet has a center frequency of Hz, use π − , π + as
2π 2 2 2 2
d 1
the interval. Then solve Δ < εx for d to get d > for the bump wavelet.
1+d 4

To show this separation requirement for the bump wavelet, consider a signal composed of
cos(2π(0.1t)) + sin((2π(0.2t)). Using the bump wavelet to obtain the CWT, the instantaneous phase of
the cosine is ϕ1(t) = 0.1t, and the instantaneous frequency is the first derivative, 0.1. Likewise, for the
1
sine component, the instantaneous frequency is 0.2. The separation inequality, 0.1 ≥ 0.3 , is true.
4
Therefore, the two signal components are IMT functions and are separated enough to use the
synchrosqueezed transform.

If you use higher frequencies, such as 0.3 and 0.4 for the instantaneous frequencies, the inequality is
1
0.1 ≥ 0.7 , which is not true. Because these signal components are not well-separated IMTs the
4
signal, cos(2π(0.3t)) + sin((2π(0.4t)), is not appropriate for use with the synchrosqueezed transform.

1-81
1 Getting Started with Wavelet Toolbox Software

Examples
CWT vs Synchrosqueezed Transform Smearing

Comparing the CWT with the synchrosqueezed transform of a quadratic chirp shows reduced energy
smearing for the synchrosqueezed transform result.

load quadchirp;
Fs = 1000;
[wt,f] = cwt(quadchirp,'bump',Fs);
subplot(2,1,1)
hp = pcolor(tquad,f,abs(wt));
hp.EdgeColor = 'none';
xlabel('Time (secs)')
ylabel('Frequency (Hz)')
title('CWT of Quadratic Chirp')
subplot(2,1,2)
wsst(quadchirp,Fs,'bump')

Low-Frequency vs. High-Frequency Component Separation

This example shows the separation needed between signal components to obtain usable results from
the synchrosqueezed transform. The signal components are 0.025, 0.05, 0.20, and 0.225 cycles per
sample. The high-frequency components, 0.20 and 0.225, do not have enough separation, so you
cannot express the whole signal as a sum of well-separated IMTs.

1-82
 Wavelet Synchrosqueezing

Define the signal and plot the synchrosqueezed components.

t = 0:2000;
x1 = cos(2*pi*.025*t);
x2 = cos(2*pi*.05*t);
x3 = cos(2*pi*.20*t);
x4 = cos(2*pi*.225*t);
x =x1+x2+x3+x4;
[sst,f] = wsst(x);
contour(t,f,abs(sst))
xlabel('Time')
ylabel('Normalized Frequency')
title('Inadequate High-Frequency Separation')

Increase the separation of the high-frequency components, and then plot the synchrosqueezed
components again.

x4 = cos(2*pi*.3*t);
x =x1+x2+x3+x4;
[sst,f] = wsst(x);
figure
contour(t,f,abs(sst))
xlabel('Time')
ylabel('Normalized Frequency')
title('Adequate High-Frequency Separation')

1-83
1 Getting Started with Wavelet Toolbox Software

All the signal components are now well-separated IMTs and are separated enough to distinguish from
each other. This signal is appropriate for use with the synchrosqueezing algorithm.

Region With Inadequate Separation

This example shows a signal with two linear chirps. A linear chirp is defined as

mt2
f (t) = cos ϕ + 2π f 0t + .
2

Its first derivative, f 0 + mt, defines the instantaneous frequency line. Use the bump wavelet and its
separation constant of 0.25. To determine the region where the two chirp signals with instantaneous
frequencies of 0.4 and 0.1 cycles per sample are not separated enough, solve this equation:

|y1 − y2 | = 0 . 25 | y1 + y2 | .

−0 . 15 0 . 15
y1 = x + 0 . 4 and y2 = x + 0 . 1 are the instantaneous frequency lines of the chirps.
1000 1000

t = 0:2000;
y1 = chirp(t,0.4,1000,0.25);
y2 = chirp(t,0.1,1000,0.25);
y = y1+y2;
wsst(y,'bump')
xlabel('Samples')

1-84
 Wavelet Synchrosqueezing

h1 = line([583 583], [0 0.5]);

h2 = line([1417 1417], [0 0.5]);
h1.Color='white';
h2.Color='white';

The vertical lines are the bounds of the region. They indicate that not enough separation occurs at
sample 583 and sample 1417. In the region between the vertical lines, the signal does not consist of
well-separated IMTs. In the regions outside the vertical lines, the signal has well-separated IMTs. You
can obtain good results from the synchrosqueezed transform in these regions.

References
[1] Mallet, S. A Wavelet Tour of Signal Processing. San Diego, CA: Academic Press, 2008, p. 89.

[2] Daubechies, I., J. Lu, and H. T. Wu. "Synchrosqueezed Wavelet Transforms: an empirical mode
decomposition-like tool." Applied and Computational Harmonic Analysis. Vol. 30(2), pp. 243–
261.

[3] Thakur, G., E. Brevdo, N. S. Fučkar, and H. T. Wu. "The synchrosqueezing algorithm for time-
varying spectral analysis: robustness properties and new paleoclimate applications." Signal
Processing. Vol. 93, pp. 1079–1094.

1-85
1 Getting Started with Wavelet Toolbox Software

See Also

Related Examples
• “Time-Frequency Reassignment and Mode Extraction with Synchrosqueezing”

1-86
 Introduction to Wavelet Families

Introduction to Wavelet Families

Several families of wavelets that have proven to be especially useful are included in this toolbox.
What follows is an introduction to some wavelet families.

• “Haar” on page 1-87

• “Daubechies” on page 1-87
• “Biorthogonal” on page 1-88
• “Coiflets” on page 1-89
• “Symlets” on page 1-90
• “Morlet” on page 1-90
• “Mexican Hat” on page 1-91
• “Meyer” on page 1-91
• “Other Real Wavelets” on page 1-91
• “Complex Wavelets” on page 1-92

To explore all wavelet families on your own, check out the Wavelet Display tool:

1 Type wvdtool at the MATLAB command line. The Wavelet Display tool appears.
2 Select a family from the Wavelet drop-down list at the top right of the tool.
3 Click the Display button. Pictures of the wavelets and their associated filters appear.
4 Obtain more information by clicking the information buttons located at the right.

Haar
Any discussion of wavelets begins with Haar wavelet, the first and simplest. The Haar wavelet is
discontinuous, and resembles a step function. It represents the same wavelet as Daubechies db1.

Daubechies
Ingrid Daubechies, one of the brightest stars in the world of wavelet research, invented what are
called compactly supported orthonormal wavelets — thus making discrete wavelet analysis
practicable.

The names of the Daubechies family wavelets are written dbN, where N is the order, and db the
“surname” of the wavelet. The db1 wavelet, as mentioned above, is the same as Haar wavelet. Here
are the wavelet functions psi of the next nine members of the family:

1-87
1 Getting Started with Wavelet Toolbox Software

You can obtain a survey of the main properties of this family by typing waveinfo('db') from the
MATLAB command line. See “Daubechies Wavelets: dbN” in the Wavelet Toolbox User's Guide for
more detail.

Biorthogonal
This family of wavelets exhibits the property of linear phase, which is needed for signal and image
reconstruction. By using two wavelets, one for decomposition (on the left side) and the other for
reconstruction (on the right side) instead of the same single one, interesting properties are derived.

1-88
 Introduction to Wavelet Families

You can obtain a survey of the main properties of this family by typing waveinfo('bior') from the
MATLAB command line. See “Biorthogonal Wavelet Pairs: biorNr.Nd” in the Wavelet Toolbox User's
Guide for more detail.

Coiflets
Built by I. Daubechies at the request of R. Coifman. The wavelet function has 2N moments equal to 0
and the scaling function has 2N-1 moments equal to 0. The two functions have a support of length
6N-1. You can obtain a survey of the main properties of this family by typing waveinfo('coif')
from the MATLAB command line. See “Coiflet Wavelets: coifN” in the Wavelet Toolbox User's Guide
for more detail.

1-89
1 Getting Started with Wavelet Toolbox Software

Symlets
The symlets are nearly symmetrical wavelets proposed by Daubechies as modifications to the db
family. The properties of the two wavelet families are similar. Here are the wavelet functions psi.

You can obtain a survey of the main properties of this family by typing waveinfo('sym') from the
MATLAB command line. See “Symlet Wavelets: symN” in the Wavelet Toolbox User's Guide for more
detail.

Morlet
This wavelet has no scaling function, but is explicit.

You can obtain a survey of the main properties of this family by typing waveinfo('morl') from the
MATLAB command line. See “Morlet Wavelet: morl” in the Wavelet Toolbox User's Guide for more
detail.

1-90
 Introduction to Wavelet Families

Mexican Hat
This wavelet has no scaling function and is derived from a function that is proportional to the second
derivative function of the Gaussian probability density function. It is also knows as the Ricker
wavelet.

You can obtain a survey of the main properties of this family by typing waveinfo('mexh') from the
MATLAB command line. See “Mexican Hat Wavelet: mexh” in the Wavelet Toolbox User's Guide for
more information.

Meyer
The Meyer wavelet and scaling function are defined in the frequency domain.

You can obtain a survey of the main properties of this family by typing waveinfo('meyer') from the
MATLAB command line. See “Meyer Wavelet: meyr” in the Wavelet Toolbox User's Guide for more
detail.

Other Real Wavelets

Some other real wavelets are available in the toolbox:

• Reverse Biorthogonal
• Gaussian derivatives family
• FIR based approximation of the Meyer wavelet

See “Additional Real Wavelets” in the Wavelet Toolbox User's Guide for more information.

1-91
1 Getting Started with Wavelet Toolbox Software

Complex Wavelets
Some complex wavelet families are available in the toolbox:

• Gaussian derivatives
• Morlet
• Frequency B-Spline
• Shannon

See “Complex Wavelets” in the Wavelet Toolbox User's Guide for more information.

1-92
 References

References
[Abr97] Abry, P. (1997), Ondelettes et turbulence. Multirésolutions, algorithmes de décomposition,
invariance d'échelles, Diderot Editeur, Paris.

[Abr03]Abry, P., P. Flandrin, M.S. Taqqu, D. Veitch (2003), “Self-similarity and long-range dependence
through the wavelet lens,” Theory and applications of long-range dependence, Birkhäuser, pp. 527–
556.

[Ant94] Antoniadis, A. (1994), “Smoothing noisy data with coiflets,” Statistica Sinica 4 (2), pp. 651–
678.

[AntO95] Antoniadis, A., G. Oppenheim, Eds. (1995), Wavelets and statistics, Lecture Notes in
Statistics 103, Springer Verlag.

[AntP98] Antoniadis A., D.T. Pham (1998), “Wavelet regression for random or irregular design,”
Comp. Stat. and Data Analysis, 28, pp. 353–369.

[AntG99] Antoniadis, A., G. Gregoire (1999), “Density and Hazard rate estimation for right-censored
data using wavelet methods,” J. R. Statist. Soc. B, 61, 1, pp. 63–84.

[ArnABEM95]Arneodo, A., F. Argoul, E. Bacry, J. Elezgaray, J.F. Muzy (1995), Ondelettes,

multifractales et turbulence, Diderot Editeur, Paris.

[Bak95] Bakshi, B. (1998), “Multiscale PCA with application to MSPC monitoring,” AIChE J. 44, pp.
1596–1610.

[BarJM03] Bardet, J.-M., G. Lang, G. Oppenheim, A. Philippe, S. Stoev, M.S. Taqqu (2003),
“Generators of long-range dependence processes: a survey” Theory and applications of long-range
dependence, Birkhäuser Boston, pp. 579–623.

[BirM97] Birgé, L., P. Massart (1997), “From model selection to adaptive estimation,” in D. Pollard
(ed.), Festchrift for L. Le Cam, Springer, pp. 55–88.

[Bri95] Brislawn, C.M. (1995), “Fingerprints to digital,” Notices of the AMS. Vol. 42, pp. 1278–1283.

[Bur96] Burke Hubbard, B. (1996), The world according to wavelets, AK Peters, Wellesley. The
French original version is titled Ondes et Ondelettes. La saga d'un outil mathématique, Pour la
Science, (1995).

[Chr06] Christophe, E., C. Mailhes, P. Duhamel (2006), “Adaptation of zerotrees using signed binary
digit representations for 3 dimensional image coding,” EURASIP Journal on Image and Video
Processing, 2007, to appear in the special issue on Wavelets in Source Coding, Communications, and
Networks, Paper ID 54679.

[Chu92a] Chui, C.K. (1992a), Wavelets: a tutorial in theory and applications, Academic Press.

[Chu92b] Chui, C.K. (1992b), An introduction to wavelets, Academic Press.

[Coh92] Cohen, A. (1992), “Ondelettes, analyses multirésolution et traitement numérique du signal,”

Ph.D. thesis, University of Paris IX, Dauphine.

[Coh95] Cohen, A. (1995), Wavelets and multiscale signal processing, Chapman and Hall.

1-93
1 Getting Started with Wavelet Toolbox Software

[CohDF92] Cohen, A., I. Daubechies, J.C. Feauveau (1992), “Biorthogonal basis of compactly
supported wavelets,” Comm. Pure Appli. Math. , vol. 45, pp. 485–560.

[CohDJV93] Cohen, A., I. Daubechies, B. Jawerth, P. Vial (1993), “Multiresolution analysis, wavelets
and fast wavelet transform on an interval,” CRAS Paris, Ser. A, t. 316, pp. 417–421.

[CoiD95] Coifman, R.R., D.L. Donoho (1995), “Translation invariant de-noising,” Lecture Notes in
Statistics, 103, pp. 125–150.

[CoiMW92] Coifman, R.R., Y. Meyer, M.V. Wickerhauser (1992), “Wavelet analysis and signal
processing,” in Wavelets and their applications, M.B. Ruskai et al. (Eds.), pp. 153–178, Jones and
Bartlett.

[CoiW92] Coifman, R.R., M.V Wickerhauser (1992), “Entropy-based algorithms for best basis
selection,” IEEE Trans. on Inf. Theory, vol. 38, 2, pp. 713–718.

[Dau92] Daubechies, I. (1992), Ten lectures on wavelets, SIAM.

[DevJL92] DeVore, R.A., B. Jawerth, B.J. Lucier (1992), “Image compression through wavelet
transform coding,” IEEE Trans. on Inf. Theory, vol. 38, 2, pp. 719–746.

[Don93] Donoho, D.L. (1993), “Progress in wavelet analysis and WVD: a ten minute tour,” in
Progress in wavelet analysis and applications, Y. Meyer, S. Roques, pp. 109–128. Frontières Ed.

[Don95] Donoho, D.L. (1995), “De-Noising by soft-thresholding,” IEEE Trans. on Inf. Theory, vol. 41,
3, pp. 613–627.

[DonJ94a] Donoho, D.L., I.M. Johnstone (1994),“Ideal spatial adaptation by wavelet shrinkage,”
Biometrika, vol. 81, pp. 425–455.

[DonJ94b] Donoho, D.L., I.M. Johnstone (1994), “Ideal de-noising in an orthonormal basis chosen
from a library of bases,” CRAS Paris, Ser I, t. 319, pp. 1317–1322.

[DonJKP95] Donoho, D.L., I.M. Johnstone, G. Kerkyacharian, D. Picard (1995), “Wavelet shrinkage:
asymptopia,” Jour. Roy. Stat. Soc., series B, vol. 57, no. 2, pp. 301–369.

[DonJKP96] Donoho, D.L., I.M. Johnstone, G. Kerkyacharian, D. Picard (1996), “Density estimation
by wavelet thresholding,” Annals of Stat., 24, pp. 508–539.

[Fla92] Flandrin, P. (1992), “Wavelet analysis and synthesis of fractional Brownian motion,” IEEE
Trans. on Inf. Th., 38, pp. 910–917.

[HalPKP97] Hall, P., S. Penev, G. Kerkyacharian, D. Picard (1997), “Numerical performance of block
thresholded wavelet estimators,” Stat. and Computing, 7, pp. 115–124.

[HarKPT98] Hardle, W., G. Kerkyacharian, D. Picard, A. Tsybakov (1998), Wavelets, approximation

and statistical applications, Lecture Notes in Statistics, 129, Springer Verlag.

[Ist94] Istas, J., G. Lang (1994), “Quadratic variations and estimation of the local Hölder index of a
Gaussian process,” Ann. Inst. Poincaré, 33, pp. 407–436.

[KahL95] Kahane, J.P., P.G Lemarié (1995), Fourier series and wavelets, Gordon and Research
Publishers, Studies in the Development of Modern Mathematics, vol 3.

[Kai94] Kaiser, G. (1994), A friendly guide to wavelets, Birkhäuser.

1-94
 References

[Lav99] Lavielle, M. (1999), “Detection of multiple changes in a sequence of dependent variables,”

Stoch. Proc. and their Applications, 83, 2, pp. 79–102.

[Lem90] Lemarié, P.G., Ed., (1990), Les ondelettes en 1989, Lecture Notes in Mathematics, Springer
Verlag.

[Mal89] Mallat, S. (1989), “A theory for multiresolution signal decomposition: the wavelet
representation,” IEEE Pattern Anal. and Machine Intell., vol. 11, no. 7, pp. 674–693.

[Mal98] Mallat, S. (1998), A wavelet tour of signal processing, Academic Press.

[Mey90] Meyer, Y. (1990), Ondelettes et opérateurs, Tome 1, Hermann Ed. (English translation:
Wavelets and operators), Cambridge Univ. Press, 1993.

[Mey93] Meyer, Y. (1993), Les ondelettes. Algorithmes et applications, Colin Ed., Paris, 2nd edition.
(English translation: Wavelets: algorithms and applications, SIAM).

[MeyR93] Meyer, Y., S. Roques, Eds. (1993), Progress in wavelet analysis and applications,
Frontières Ed.

[MisMOP93a] Misiti, M., Y. Misiti, G. Oppenheim, J.M. Poggi (1993a), “Analyse de signaux classiques
par décomposition en ondelettes,” Revue de Statistique Appliquée, vol. XLI, no. 4, pp. 5–32.

[MisMOP93b] Misiti, M., Y. Misiti, G. Oppenheim, J.M. Poggi (1993b), “Ondelettes en statistique et
traitement du signal,” Revue de Statistique Appliquée, vol. XLI, no. 4, pp. 33–43.

[MisMOP94] Misiti, M., Y. Misiti, G. Oppenheim, J.M. Poggi (1994), “Décomposition en ondelettes et
méthodes comparatives: étude d'une courbe de charge électrique,” Revue de Statistique Appliquée,
vol. XLII, no. 2, pp. 57–77.

[MisMOP03] Misiti, M., Y. Misiti, G. Oppenheim, J.-M. Poggi (2003), “Les ondelettes et leurs
applications,” Hermes.

[MisMOP07] Misiti, M., Y. Misiti, G. Oppenheim, J.-M. Poggi (2007), Wavelets and their applications,
ISTE DSP Series.

[NasS95] Nason, G.P., B.W. Silverman (1995), “The stationary wavelet transform and some statistical
applications,” Lecture Notes in Statistics, 103, pp. 281–299.

[Ogd97] Ogden, R.T. (1997), Essential wavelets for statistical applications and data analysis,
Birkhäuser.

[PesKC96] Pesquet, J.C., H. Krim, H. Carfatan (1996), “Time-invariant orthonormal wavelet

representations,” IEEE Trans. Sign. Proc., vol. 44, 8, pp. 1964–1970.

[Sai96] Said A., W.A. Pearlman (1996), “A new, fast, and efficient image codec based on set
partitioning in hierarchical trees,” IEEE Trans. on Circuits and Systems for Video Technology, Vol. 6,
No. 3, pp. 243–250.

[Sha93] Shapiro J.M. (1993), “Embedded image coding using zerotrees of wavelet coefficients,” IEEE
Trans. Signal Proc., Vol. 41, No. 12, pp. 3445–3462.

[StrN96] Strang, G., T. Nguyen (1996), Wavelets and filter banks, Wellesley-Cambridge Press.

[Swe98] Sweldens, W. (1998), “The Lifting Scheme: a Construction of Second Generation of

Wavelets,” SIAM J. Math. Anal., 29 (2), pp. 511–546.

1-95
1 Getting Started with Wavelet Toolbox Software

[Teo98] Teolis, A. (1998), Computational signal processing with wavelets, Birkhäuser.

[VetK95] Vetterli, M., J. Kovacevic (1995), Wavelets and subband coding, Prentice Hall.

[Wal99] Walker, J.S. (1999), “Wavelet-Based Image Compression,” University of Wisconsin, Eau
Claire, Wisconsin, USA, , Sub-chapter of CRC Press book: Transform and Data Compression. A Primer
on Wavelets and Their Scientific Applications. A second edition is published in 2008.

[Wic91] Wickerhauser, M.V. (1991), “INRIA lectures on wavelet packet algorithms,” Proceedings
ondelettes et paquets d'ondes, 17–21 June, Rocquencourt France, pp. 31–99.

[Wic94] Wickerhauser, M.V. (1994), Adapted wavelet analysis from theory to software algorithms,
A.K. Peters.

[Zee98] Zeeuw, P.M. (1998), “Wavelet and image fusion,” CWI, Amsterdam, March 1998, http:/
www.cwi.nl/~pauldz/

1-96
 2

Using Wavelets

This chapter takes you step-by-step through examples that teach you how to use the graphical tools
and command-line functions.

• “Introduction to Wavelet Toolbox App and Functions” on page 2-2

• “Wavelets: Working with Images” on page 2-3
• “Density Estimation Using Wavelets” on page 2-8
• “1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App” on page 2-13
• “2-D Wavelet Coefficient Selection Using the Wavelet Analyzer App” on page 2-20
• “1-D Extension Using the Command Line” on page 2-25
• “2-D Extension Using the Command Line” on page 2-26
• “Image Fusion” on page 2-27
• “1-D Fractional Brownian Motion Synthesis” on page 2-33
• “New Wavelet for CWT” on page 2-35
• “Additional Wavelet GUIs” on page 2-38
2 Using Wavelets

Introduction to Wavelet Toolbox App and Functions

Wavelet Toolbox software contains graphical tools and command-line functions that let you

• Examine and explore properties of individual wavelets and wavelet packets

• Examine statistics of signals and signal components
• Perform a continuous wavelet transform of a 1-D signal
• Perform discrete analysis and synthesis of 1-D and 2-D signals
• Perform wavelet packet analysis of 1-D and 2-D signals
• Compress and remove noise from signals and images

In addition to the above, the toolbox makes it easy to customize the presentation and visualization of
your data. You choose

• Which signals to display

• A region of interest to magnify
• A coloring scheme for display of wavelet coefficient details

Note All the Wavelet Analyzer app tools described here let you import information from and
export information to either the disk or workspace.

2-2
 Wavelets: Working with Images

Wavelets: Working with Images

This section provides additional information about working with images in the Wavelet Toolbox
software. It describes the types of supported images and how the MATLAB environment represents
them, as well as techniques for analyzing color images.

Understanding Images in the MATLAB Environment

The basic data structure in MATLAB is the rectangular matrix, an ordered set of real or complex
elements. This object is naturally suited to the representation of images, which are real-valued,
ordered sets of color or intensity data. (This toolbox does not support complex-valued images.)

The word pixel is derived from picture element and usually denotes a single dot on a computer
display, or a single element in an image matrix. You can select a single pixel from an image matrix
using normal matrix subscripting. For example:

I(2,15)

returns the value of the pixel at row 2 and column 15 of the image I. By default, MATLAB scales
images to fill the display axes; therefore, an image pixel may use more than a single pixel on the
screen.

Indexed Images
A typical color image requires two matrices: a colormap and an image matrix. The colormap is an
ordered set of values that represent the colors in the image. For each image pixel, the image matrix
contains a corresponding index into the colormap. (The elements of the image matrix are floating-
point integers, or flints, which MATLAB stores as double-precision values.)

The size of the colormap matrix is n-by-3 for an image containing n colors. Each row of the colormap
matrix is a 1-by-3 red, green, blue (RGB) color vector

color = [R G B]

that specifies the intensity of the red, green, and blue components of that color. R, G, and B are real
scalars that range from 0.0 (black) to 1.0 (full intensity). MATLAB translates these values into display
intensities when you display an image and its colormap.

When MATLAB displays an indexed image, it uses the values in the image matrix to look up the
desired color in the colormap. For instance, if the image matrix contains the value 18 in matrix
location (86,198), the color for pixel (86,198) is the color from row 18 of the colormap.

2-3
2 Using Wavelets

Outside MATLAB, indexed images with n colors often contain values from 0 to n–1. These values are
indices into a colormap with 0 as its first index. Since MATLAB matrices start with index 1, you must
increment each value in the image, or shift up the image, to create an image that you can manipulate
with toolbox functions.

Wavelet Decomposition of Indexed Images

Indexed images can be thought of as scaled intensity images, with matrix elements containing only
integers from 1 to n, where n is the number of discrete shades in the image.

If the colormap is not provided, the Wavelet Analyzer app displays the image and processing results
using a monotonic colormap with max(max(X))-min(min(X))+1 colors.

Since the image colormap is only used for display purposes, some indexed images may need to be
preprocessed to achieve the correct results from the wavelet decomposition.

In general, color indexed images do not have linear, monotonic colormaps and need to be converted
to the appropriate gray-scale indexed image before performing a wavelet decomposition.

How Decompositions Are Displayed

Note that the coefficients, approximations, and details produced by wavelet decomposition are not
indexed image matrices.

To display these images in a suitable way, the Wavelet Analyzer app follows these rules:

• Reconstructed approximations are displayed using the colormap map.

• The coefficients and the reconstructed details are displayed using the colormap map applied to a
rescaled version of the matrices.

2-4
 Wavelets: Working with Images

RGB (Truecolor) Images

An RGB image, sometimes referred to as a truecolor image, is stored in MATLAB as an m-by-n-by-3
data array that defines red, green, and blue color components for each individual pixel. RGB images
do not use a palette. The color of each pixel is determined by the combination of the red, green, and
blue intensities stored in each color plane at the pixel's location. Graphics file formats store RGB
images as 24-bit images, where the red, green, and blue components are 8 bits each. This yields a
potential of 16 million colors.

The precision with which a real-life image can be replicated led to the nickname “truecolor image.”
An RGB MATLAB array can be of class double, single, uint8, or uint16. In an RGB array of class
double, each color component is a value between 0 and 1.

The color components of an 8-bit RGB image are integers in the range [0, 255] rather than floating-
point values in the range [0, 1].

Wavelet Decomposition of Truecolor Images

The truecolor images analyzed are m-by-n-by-3 arrays of uint8. Each of the three-color components
is a matrix that is decomposed using the 2-D wavelet decomposition scheme.

Other Images
Wavelet Toolbox software lets you work with some other types of images. Using the imread function,
the various tools using images try to load indexed images from files that are not MAT files (for
example, PCX files).

These tools are:

• 2-D Discrete Wavelet Analysis

• 2-D Wavelet Packet Analysis
• 2-D Stationary Wavelet Analysis

For more information on the supported file types, type help imread.

Use the imfinfo function to find the type of image stored in the file. If the file does not contain an
indexed image, the load operation fails.

Image Conversion
Image Processing Toolbox software provides a comprehensive set of functions that let you easily
convert between image types. If you do not have Image Processing Toolbox software, the examples
below demonstrate how this conversion may be performed using basic MATLAB commands.

Example 1: Converting Color Indexed Images

load xpmndrll
whos

Name Size Bytes Class

X2 192x200 307200 double array

2-5
2 Using Wavelets

Name Size Bytes Class

map 64x3 1536 double array

image(X2)
title('Original Color Indexed Image')
colormap(map); colorbar

The color bar to the right of the image is not smooth and does not monotonically progress from dark
to light. This type of indexed image is not suitable for direct wavelet decomposition with the toolbox
and needs to be preprocessed.

First, separate the color indexed image into its RGB components:

R = map(X2,1); R = reshape(R,size(X2));
G = map(X2,2); G = reshape(G,size(X2));
B = map(X2,3); B = reshape(B,size(X2));

Next, convert the RGB matrices into a gray-scale intensity image, using the standard perceptual
weightings for the three-color components:

Xrgb = 0.2990*R + 0.5870*G + 0.1140*B;

Then, convert the gray-scale intensity image back to a gray-scale indexed image with 64 distinct
levels and create a new colormap with 64 levels of gray:

n = 64; % Number of shades in new indexed image

X = round(Xrgb*(n-1)) + 1;
map2 = gray(n);
figure
image(X), title('Processed
Gray Scale Indexed Image')
colormap(map2), colorbar

2-6
 Wavelets: Working with Images

The color bar of the converted image is now linear and has a smooth transition from dark to light. The
image is now suitable for wavelet decomposition.

Finally, save the converted image in a form compatible with the Wavelet Toolbox Wavelet Analyzer
app:

baboon = X;
map = map2;
save baboon baboon map

Example 2: Converting an RGB TIF Image

Suppose the file myImage.tif contains an RGB image (noncompressed) of size S1xS2. Use the
following commands to convert this image:

A = imread('myImage.tif');
% A is an S1xS2x3 array of uint8.

A = double(A);
Xrgb = 0.2990*A(:,:,1) + 0.5870*A(:,:,2) + 0.1140*A(:,:,3);
NbColors = 255;
X = wcodemat(Xrgb,NbColors);
map = pink(NbColors);

The same program can be used to convert BMP or JPEG files.

2-7
2 Using Wavelets

Density Estimation Using Wavelets

This section takes you through the features of 1-D wavelet density estimation using one of the
Wavelet Toolbox specialized tools.

The toolbox provides Wavelet Analyzer app to estimate the density of a sample and complement well
known tools like the histogram (available from the MATLAB core) or kernel based estimates.

For the examples in this section, switch the extension mode to symmetric padding, using the
command

dwtmode('sym')

1-D Estimation Using the Wavelet Analyzer App

1 Start the Density Estimation 1-D Tool.

From the MATLAB prompt, type waveletAnalyzer.

The Wavelet Analyzer appears.

Click the Density Estimation 1-D menu item. The discrete wavelet analysis tool for 1-D density
estimation appears.

2-8
 Density Estimation Using Wavelets

2 Load data.

At the MATLAB command line, type

load ex1cusp1

In the Density Estimation 1-D tool, choose File > Import from Workspace.

When the Import from Workspace dialog box appears, select ex1cusp1. Click OK to import
the noisy cusp data.

The sample, a 64-bin histogram, and the processed data obtained after a binning are displayed.
In this example, we'll accept the default value for the number of bins (250). The binned data,
suitably normalized, will be processed by wavelet decomposition.
3 Perform a Wavelet Decomposition of the binned data.

Select the sym6 wavelet from the Wavelet menu and select 4 from the Level menu, and click the
Decompose button. After a pause for computation, the tool displays the detail coefficients of the
decomposition of the binned data.
4 Perform a density estimation.

Accept the defaults of global soft thresholding. The sliders located on the right of the window
control the level dependent thresholds, indicated by dashed blue lines running horizontally
through the graphs on the left of the window.

Continue by clicking the Estimate button.

2-9
2 Using Wavelets

You can see that the estimation process delivers a very irregular resulting density. The density
estimate (in purple) is the normalized sum of the signals located below it: the approximation a4
and the reconstructed details after coefficient thresholding.
5 Perform thresholding.

You can experiment with the various predefined thresholding strategies by selecting the
appropriate options from the menu located on the right of the window or directly by dragging the
dashed blue lines with the left mouse button. Let's try another estimation method.

From the menu Select thresholding method, select the item By level threshold 2. Next, click
the Estimate button.

2-10
 Density Estimation Using Wavelets

The estimated density is more satisfactory. It correctly identifies the smooth part of the density
and the cusp at 0.7.

Importing and Exporting Information from the Wavelet Analyzer App

The tool lets you save the estimated density to disk. The toolbox creates a MAT-file in the current
folder with a name you choose.

To save the estimated density, use the menu option File > Save Density. A dialog box appears that
lets you specify a folder and filename for storing the density. Type the name dex1cusp. After saving
the density data to the file dex1cusp.mat, load the variables into your workspace:

load dex1cusp
whos

Name Size Bytes Class

ex1cusp1 1x1000 8000 double array
thrParams 1x4 544 cell array
wname 1x4 8 char array
xdata 1x250 2000 double array
ydata 1x250 2000 double array

2-11
2 Using Wavelets

The original noisy cusp data ex1cusp1 has 1000 samples. The variables thrParams, wname, xdata,
and ydata are stored in dex1cusp.mat. The estimated density is given by xdata and ydata. The
length of these vectors is equal to the number of bins you choose in step 4. In addition, the
parameters of the estimation process are given by the wavelet name in wname.

wname

wname =
sym6

and the level dependent thresholds contained in thrParams, which is a cell array of length 4 (the
level of the decomposition). For i from 1 to 4, thrParams{i} contains the lower and upper bounds of
the interval of thresholding and the threshold value (since interval dependent thresholds are
allowed). For more information, see “1-D Adaptive Thresholding of Wavelet Coefficients”. For
example, for level 1,

thrParams{1}
ans =
0.0560 0.9870 2.1179

Note When you load data from a file using the menu option File > Load Data for Density
Estimate, the first 1-D variable encountered in the file is considered the signal. Variables are
inspected in alphabetical order.

At the end of this section, turn the extension mode back to zero padding using

dwtmode('zpd')

2-12
 1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

1-D Wavelet Coefficient Selection Using the Wavelet Analyzer

App
This section takes you through the features of 1-D selection of wavelet coefficients using one of the
Wavelet Toolbox specialized tools. The toolbox provides the Wavelet Analyzer app to explore some
reconstruction schemes based on various wavelet coefficients selection strategies:

• Global selection of biggest coefficients (in absolute value)

• By level selection of biggest coefficients
• Automatic selection of biggest coefficients
• Manual selection of coefficients

For this section, switch the extension mode to symmetric padding using the command

dwtmode('sym')

1 Start the Wavelet Coefficients Selection 1-D Tool.

From the MATLAB prompt, type waveletAnalyzer.

The Wavelet Analyzer appears.

Click the Wavelet Coefficients Selection 1-D menu item. The discrete wavelet coefficients
selection tool for 1-D signals appears.

2-13
2 Using Wavelets

2 Load data.

At the MATLAB command prompt, type

load noisbump

In the Wavelet Coefficients Selection 1-D tool, select File > Import Signal from
Workspace. When the Import from Workspace dialog box appears, select the noisbump
variable. Click OK to import the noisy bumps data
3 Perform a Wavelet Decomposition.

Select the db3 wavelet from the Wavelet menu and select 6 from the Level menu, and then click
the Analyze button.

The tool displays below the original signal (on the left) its wavelet decomposition: the
approximation coefficients A6 and detail coefficients from D6 at the top to D1 at the bottom. In
the middle of the window, below the synthesized signal (which at this step is the same, since all
the wavelet coefficients are kept) it displays the selected coefficients.

Selecting Biggest Coefficients Globally

On the right of the window, find a column labeled Kept. The last line shows the total number of
coefficients: 1049. This is a little bit more than the number of observations, which is 1024. You
can choose the number of selected biggest coefficients by typing a number instead of 1049 or by
using the slider. Type 40 and press Enter. The numbers of selected biggest coefficients level by

2-14
 1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

level are updated (but cannot be modified since Global is the current selection method). Then
click the Apply button. The resulting coefficients are now displayed.

In the previous trial, the approximation coefficients were all kept. It is possible to relax this
constraint by selecting another option from the App. cfs menu (Approximation Coefficients
abbreviation). Choose the Unselect option and click the Apply button.

None of the approximation coefficients are kept.

From the App. cfs menu, select the Selectable option. Type 80 for the number of selected
biggest coefficients and press Enter. Then, click the Apply button.

2-15
2 Using Wavelets

Some of the approximation coefficients (15) have been kept.

Selecting Biggest Coefficients by Level

From the Define Selection method menu, select the By Level option. You can choose the
number of selected biggest coefficients by level or select it using the sliders. Type 4 for the
approximation and each detail, and then click the Apply button.

Selecting Coefficients Manually

From the Define Selection method menu, select the Manual option. The tool displays on the
left part, below the original signal, its wavelet decomposition. At the beginning, no coefficients
are kept so no selected coefficient is visible and the synthesized signal is null.

2-16
 1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

Select 16 coefficients individually by double clicking each of them using the left mouse button.
The color of selected coefficients switches from green to yellow for the details and from blue to
yellow for the approximation, which appear on the left of the window and appear in yellow on the
middle part. Click the Apply button.

You can deselect the currently selected coefficients by double clicking each of them. Another way
to select or deselect a set of coefficients is to use the selection box. Drag a rubber band box (hold
down the left mouse button) over a portion of the coefficient axes (original or selected)
containing all the currently selected coefficients. Click the Unselect button located on the right
of the window. Click the Apply button. The tool displays the null signal again.

Note that when the coefficients are very close, it is easier to zoom in before selecting or
deselecting them.

Drag a rubber band box over the portion of the coefficient axes around the position 800 and
containing all scales and click the Select button. Click the Apply button.

2-17
2 Using Wavelets

This illustrates that wavelet analysis is a local analysis since the signal is perfectly reconstructed
around the position 800. Check the Show Original Signal to magnify it.

Selecting Coefficients Automatically

From the Define Selection method menu, select the Stepwise movie option. The tool displays
the same initial window as in the manual selection mode, except for the left part of it.

Let's perform the stepwise movie using the k biggest coefficients, from k = 1 to k = 31 in steps
of 1, click the Start button. As soon as the result is satisfactory, click the Stop button.

4 Save the synthesized signal.

The tool lets you save the synthesized signal to disk. The toolbox creates a MAT-file in the current
folder with a name you choose.

2-18
 1-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

To save the synthesized signal from the present selection, use the menu option File > Save
Synthesized Signal. A dialog box appears that lets you specify a folder and filename for storing
the signal and the wavelet name.

At the end of this section, turn back the extension mode to zero padding using the command

dwtmode('zpd')

2-19
2 Using Wavelets

2-D Wavelet Coefficient Selection Using the Wavelet Analyzer

App
This section takes you through the features of 2-D selection of wavelet coefficients using one of the
Wavelet Toolbox specialized tools. The toolbox provides the Wavelet Analyzer app to explore some
reconstruction schemes based on various wavelet coefficient selection strategies:

• Global selection of biggest coefficients (in absolute value)

• By level selection of biggest coefficients
• Automatic selection of biggest coefficients.

This section will be short since the functionality are similar to the 1-D ones examined in the previous
section.

For this section, switch the extension mode to symmetric padding using the command

dwtmode('sym')

1 Start the Wavelet Coefficients Selection 2-D Tool.

From the MATLAB prompt, type waveletAnalyzer.

The Wavelet Analyzer appears.

Click the Wavelet Coefficients Selection 2-D menu item. The discrete wavelet coefficients
selection tool for images appears.

2-20
 2-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

2 Load data.

At the MATLAB command prompt, type

load noiswom;

In the Wavelet Coefficients Selection 2-D tool, select File > Import from Workspace. When
the Import from Workspace dialog box appears, select the X variable. Click OK to import the
image.
3 Perform a Wavelet Decomposition.

Select the sym4 wavelet from the Wavelet menu and select 4 from the Level menu, and then
click the Analyze button.

2-21
2 Using Wavelets

The tool displays its wavelet decomposition below the original image (on the left). The selected
coefficients are displayed in the middle of the window, below the synthesized image (which, at
this step, is the same since all the wavelet coefficients are kept). There are 11874 coefficients, a
little bit more than the original image number of pixels, which is 96x96 = 9216.

Note The difference between 9216 and 11874 comes from the extra coefficients generated by
the redundant DWT using the current extension mode (symmetric, 'sym'). Because 96 is
divisible by 24 = 16, using the periodic extension mode ('per') for the DWT, you obtain for each
level the minimum number of coefficients. More precisely, if you type dwtmode('per') and
repeat steps 2 to 5, you obtain 9216 coefficients.

Selecting Biggest Coefficients Globally

On the right of the window, find a column labeled Kept. The last line shows the total number of
coefficients: 11874. This is a little bit more than the original image number of pixels. You can
choose the number of selected biggest coefficients by typing a number instead of 11874, or by
using the slider. Type 1100 and press Enter. The numbers of selected biggest coefficients level
by level are updated (but cannot be modified, since Global is the current selection method).

Then click the Apply button.

In the previous operation, all the approximation coefficients were kept. It is possible to relax this
constraint by selecting another option from the App. cfs menu (see “1-D Wavelet Coefficient
Selection Using the Wavelet Analyzer App” on page 2-13).

Selecting Biggest Coefficients by Level

2-22
 2-D Wavelet Coefficient Selection Using the Wavelet Analyzer App

Selecting Biggest Coefficients by Level. From the Define Selection method menu, select the By
Level option. You can choose the number of selected biggest coefficients by level, or select it
using the sliders. Type 100 for each detail, and then click the Apply button.

Selecting Coefficients Automatically

From the Define Selection method menu, select the Stepwise movie option. The tool displays
its wavelet decomposition on the left, below the original image. At the beginning, no coefficients
are kept so the synthesized image is null. Perform the stepwise movie using the k biggest
coefficients, from k = 144 to k = 1500, in steps of 20. Click the Start button. As soon as the
result is satisfactory, click the Stop button.

We've stopped the movie at 864 coefficients (including the number of approximation coefficients).

2-23
2 Using Wavelets

4 Save the synthesized image.

This tool lets you save the synthesized image to disk. The toolbox creates a MAT-file in the
current folder with a name you choose.

To save the synthesized image from the present selection, use the menu option File > Save
Synthesized Image. A dialog box appears that lets you specify a folder and filename for storing
the image and, in addition, the colormap and the wavelet name.

At the end of this section, turn back the extension mode to zero padding using the command

dwtmode('zpd')

2-24
 1-D Extension Using the Command Line

1-D Extension Using the Command Line

The function wextend performs signal extension. For more information, see its reference page.

2-25
2 Using Wavelets

2-D Extension Using the Command Line

The function wextend performs image extension. For more information, see its reference page.

2-26
 Image Fusion

Image Fusion
For the example, switch the extension mode to symmetric padding, using the command:

dwtmode('sym')

The toolbox requires only one function for image fusion: wfusimg. You'll find full information about
this function in its reference page. For more details on fusion methods see the wfusmat function.

In this section, you will learn how to:

• Load images
• Perform decompositions
• Merge images from their decompositions
• Restore images from their decompositions
• Save image after fusion

The principle of image fusion using wavelets is to merge the wavelet decompositions of the two
original images using fusion methods applied to approximations coefficients and details coefficients
(see [MisMOP03] and [Zee98] in “References” on page 1-93).

The two images must be of the same size and are supposed to be associated with indexed images on a
common colormap (see wextend to resize images).

Two examples are examined: the first one merges two different images leading to a new image and
the second restores an image from two fuzzy versions of an original image.

Image Fusion Using the Command Line

Example 1: Fusion of Two Different Images

1 Load two original images: a mask and a bust.

load mask; X1 = X;
load bust; X2 = X;
2 Merge the two images from wavelet decompositions at level 5 using db2 by taking two different
fusion methods: fusion by taking the mean for both approximations and details,

XFUSmean = wfusimg(X1,X2,'db2',5,'mean','mean');

and fusion by taking the maximum for approximations and the minimum for the details.

XFUSmaxmin = wfusimg(X1,X2,'db2',5,'max','min');
3 Plot original and synthesized images.

colormap(map);
subplot(221), image(X1), axis square, title('Mask')
subplot(222), image(X2), axis square, title('Bust')
subplot(223), image(XFUSmean), axis square,
title('Synthesized image, mean-mean')
subplot(224), image(XFUSmaxmin), axis square,
title('Synthesized image, max-min')

2-27
2 Using Wavelets

Example 2: Restoration by Fusion from Fuzzy Images

1 Load two fuzzy versions of an original image.

load cathe_1; X1 = X;
load cathe_2; X2 = X;
2 Merge the two images from wavelet decompositions at level 5 using sym4 by taking the
maximum of absolute value of the coefficients for both approximations and details.

XFUS = wfusimg(X1,X2,'sym4',5,'max','max');
3 Plot original and synthesized images.

colormap(map);
subplot(221), image(X1), axis square,
title('Catherine 1')
subplot(222), image(X2), axis square,
title('Catherine 2')
subplot(223), image(XFUS), axis square,
title('Synthesized image')

2-28
 Image Fusion

The synthesized image is a restored version of good quality of the common underlying original
image.

Image Fusion Using the Wavelet Analyzer App

The principle of image fusion using wavelets is to merge the wavelet decompositions of the two
original images using fusion methods applied to approximations coefficients and details coefficients
(see [MisMOP03] and [Zee98] in “References” on page 1-93).

The two images must be of the same size and are supposed to be associated with indexed images on a
common colormap (see wextend to resize images).

Two examples are examined: the first one merges two different images leading to a new image and
the second restores an image from two fuzzy versions of an original image.
1 Start the Wavelet Analyzer App.

From the MATLAB prompt, type waveletAnalyzer to display the Wavelet Analyzer and then
click the Image Fusion menu item to display the Image Fusion Tool.

2-29
2 Using Wavelets

2 Load the original images: a mask and a bust.

load mask; X1 = X;
load bust; X2 = X;

In the Image Fusion tool, select File > Load or Import Image 1 > Import from Workspace.
When the Import from Workspace dialog box appears, select the X1 variable, which loads the
mask image.

Perform the same sequence choosing the X2 variable to load the bust image.
3 Perform wavelet decompositions.

Using the Wavelet and Level menus located to the upper right, determine the wavelet family, the
wavelet type, and the number of levels to be used for the analysis.

For this analysis, select the db2 wavelet at level 5.

Click the Decompose button.

After a pause for computation, the tool displays the two analyses.

4 Merge two images from their decompositions.

From Select Fusion Method frame, select the item mean for both Approx. and Details. Next,
click the Apply button.

2-30
 Image Fusion

The synthesized image and its decomposition (which is equal to the fusion of the two
decompositions) appear. The new image produced by fusion clearly exhibits features from the
two original ones.

Let us now examine another example illustrating restoration using image fusion.
5 Restore the image using image fusion.

From the File menu, load Image 1 by selecting the MAT-file cathe_1.mat, and Image 2 by
selecting the MAT-file cathe_2.mat.
6 Using the Wavelet and Level menus, select the sym4 wavelet at level 5. Click the Decompose
button.
7 From Select Fusion Method frame, select the item max for both Approx. and Details. Next,
click the Apply button.

The synthesized image is a restored version of good quality of the common underlying original
image.

2-31
2 Using Wavelets

Saving the Synthesized Image

The Image Fusion Tool lets you save the synthesized image to disk. The toolbox creates a MAT-file in
the current folder with a name you choose.

To save the synthesized image from the present selection, use the menu option File > Save
Synthesized Image.

A dialog box appears that lets you specify a folder and filename for storing the image. After you save
the image data to the file rescathe.mat, the synthesized image is given by X and the colormap by
map.

2-32
 1-D Fractional Brownian Motion Synthesis

1-D Fractional Brownian Motion Synthesis

This example shows how to generate a fractional Brownian motion signal using the wfbm function.

A fractional Brownian motion (fBm) is a continuous-time Gaussian process depending on the Hurst
parameter 0 < H < 1. It generalizes the ordinary Brownian motion corresponding to H = 0.5 and
whose derivative is white noise. The fBm is self-similar in distribution and the variance of the
increments is given by

Var(fBm(t) - fBm(s)) = v abs(t-s)^(2H),

where v is a positive constant. The fBm exhibits long-range dependence for H > 0.5 and short or
intermediate dependence for H < 0.5.

For purposes of reproducibility, set the random seed to the default value. Generate fractional
Brownian motion with length 1000 for H = 0.3. Plot the result.

rng default
H = 0.3;
len = 1000;
fBm03 = wfbm(H,len,'plot');

Generate fractional Brownian motion with length 1000 for H = 0.7. Plot the result. Because H > 0.5,
the fractional Brownian motion exhibits a stronger low-frequency component and has, locally, less
irregular behavior.

2-33
2 Using Wavelets

rng default
H = 0.7;
fBm07 = wfbm(H,len,'plot');

Confirm the previous syntax is equivalent to generating fractional Brownian motion using the
orthogonal db10 wavelet and six reconstruction steps.

rng default
w = 'db10';
ns = 6;
fBm07x = wfbm(H,len,w,ns);
max(abs(fBm07-fBm07x))

ans = 0

2-34
 New Wavelet for CWT

New Wavelet for CWT

This example illustrates how to generate a new wavelet starting from a pattern.

The principle for designing a new wavelet for CWT is to approximate a given pattern using least
squares optimization under constraints leading to an admissible wavelet well suited for the pattern
detection using the continuous wavelet transform [1].

Load an original pattern: a pseudo sine.

load ptpssin1
who

Your variables are:

IntVAL X Y caption

The variables X and Y contain the pattern. Integrate the pattern over the interval [0, 1]. Plot the
pattern.

dX = max(diff(X));
patternInt = dX*sum(Y);
disp(['Integral of pattern = ',num2str(patternInt)]);

Integral of pattern = 0.15915

plot(X,Y)
title('Original Pattern')
grid on

2-35
2 Using Wavelets

The pattern on the interval [0, 1] integrates to 0.15915. So it is not a wavelet but it is a good
candidate since it oscillates like a wavelet.

To synthesize a new wavelet adapted to the given pattern, use a least squares polynomial
approximation of degree 6 with constraints of continuity at the beginning and the end of the pattern.

[psi,xval,nc] = pat2cwav(Y, 'polynomial',6, 'continuous');

The new wavelet is given by xval and nc*psi.

figure
plot(X,Y,'-',xval,nc*psi,'--')
grid on
legend('Original Pattern','Adapted Wavelet','Location','NorthWest')

2-36
 New Wavelet for CWT

Check that psi satisfies the definition of a wavelet by confirming that it integrates to zero and has L2
norm is equal to 1.

dxval = max(diff(xval));
newWaveletIntegral = dxval*sum(psi);
disp(['Integral of new wavelet = ',num2str(newWaveletIntegral)])

Integral of new wavelet = 1.9626e-05

newWaveletSqN = dxval*sum(psi.^2);
disp(['New wavelet has L2-norm = ',num2str(newWaveletSqN)])

New wavelet has L2-norm = 1

References
[1] Misiti, M., Y. Misiti, G. Oppenheim, and J.-M. Poggi. Les ondelettes et leurs applications. France:
Hermes Science/Lavoisier, 2003.

See Also
pat2cwav

2-37
2 Using Wavelets

Additional Wavelet GUIs

1-D Extension Using the Signal Extension Tool

1 Load data. At the MATLAB command prompt, type

load noisbloc;
2 Start the Signal Extension Tool.

From the MATLAB prompt, type sigxtool. The Signal Extension tool appears.
3 In the Signal Extension tool, select File > Import from Workspace. When the Import from
Workspace dialog box appears, select the noisbloc variable. Click OK to import the data

4 Extend the signal.

Enter 1300 in the Desired Length box of the extended signal, and select the Left option from
the Direction to extend menu. Then accept the default Symmetric for the Extension mode,
and click the Extend button.

2-38
 Additional Wavelet GUIs

The tool displays the original signal delimited by a red box and the transformed signal delimited
by a yellow box. The signal has been extended by left symmetric boundary values replication.

Select the Both option from the Direction to extend menu and select the Continuous option
from the Extension mode menu. Click the Extend button.

The signal is extended in both directions by replicating the first value to the left and the last
value to the right, respectively.

Extending Signal for SWT

Since the decomposition at level k of a signal using SWT requires that 2^k divides evenly into the
length of the signal, the tool provides a special option dedicated to this kind of extension.

Select the For SWT option from the Extension mode menu. Click the Extend button.

2-39
2 Using Wavelets

Since the signal is of length 1024 = 2^10, no extension is needed so the Extend button is
ineffective.

From the File menu, choose the Example Extension option and select the last item of the list.

Since the signal is of length 1000 and the decomposition level needed for SWT is 10, the tool
performs a minimal right periodic extension. The extended signal is of length 1024.

Select 4 from the SWT Decomposition Level menu, and then click the Extend button. The tool
performs a minimal right periodic extension leading to an extended signal of length 1008 (because
1008 is the smallest integer greater than 1000 divisible by 2^4 = 16).

Select 2 from the SWT Decomposition Level menu. Since 1000 is divisible by 4, no extension is
needed.

Truncating Signal

2-40
 Additional Wavelet GUIs

The same tool allows you to truncate a signal.

Since truncation is not allowed for the special mode For SWT, select the Periodic option from the
Extension mode menu. Type 900 for the desired length and press Enter. Click the Truncate button.

The tool displays the original signal delimited by a red box and the truncated signal delimited by a
yellow box. The signal has been truncated by deleting 100 values on the right side.

Importing and Exporting Information

This tool lets you save the transformed signal to disk. The toolbox creates a MAT-file in the current
folder with a name you choose.

To save the transformed signal, use the menu option File > Save Transformed Signal. A dialog box
appears that lets you specify a folder and filename for storing the image. Type the name tfrqbrk.
After saving the signal data to the file tfrqbrk.mat, load the variable into your workspace:

load tfrqbrk
whos

Name Size Bytes Class

tfrqbrk 1x900 7200 double array

2-D Extension Using the Image Extension Tool

1 Start the Image Extension Tool.

From the MATLAB prompt, type imgxtool. The Image Extension tool appears.
2 Extend (or truncate) the image.

From the File menu, choose the Example Extension option and select the first item of the list.

2-41
2 Using Wavelets

The tool displays the original image delimited by a red box and the transformed image delimited
by a yellow box. The image has been extended by zero padding. The right part of the window
allows you to control the parameters of the extension/truncation process for the vertical and
horizontal directions, respectively. The possibilities are similar to the 1-D ones described in “1-D
Extension Using the Command Line” on page 2-25.

Importing and Exporting Information

This tool lets you save the transformed image to disk. The toolbox creates a MAT-file in the current
folder with a name you choose.

To save the transformed image, use the menu option File > Save Transformed Image.

A dialog box appears that lets you specify a folder and filename for storing the image. Type the name
woman2. After saving the image data to the file woman2.mat, load the variable into your workspace:

load woman2
whos

Name Size Bytes Class

woman2 200x220 352000 double array
map 253x3 6120 double array

The transformed image is stored together with its colormap.

Fractional Brownian Motion Synthesis Tool

1 Start the Fractional Brownian Motion Synthesis Tool.

From the MATLAB prompt, type wfbmtool. The 1-D Fractional Brownian Motion Synthesis Tool
appears.

2-42
 Additional Wavelet GUIs

2 Generate fBm.

From the Fractal Index edit box, type 0.3 and from the Seed frame, select the item State and
set the value to 0. Next, click the Generate button.

The synthesized signal exhibits a locally highly irregular behavior.

3 Now let us try another value for the fractal index. From the Fractal Index edit box, type 0.7 and
from the Seed frame, select the item State and set the value to 0. Next, click the Generate
button.

2-43
2 Using Wavelets

The synthesized signal clearly exhibits a stronger low-frequency component and has locally a less
irregular behavior. These properties can be investigated by clicking the Statistics button.

Saving the Synthesized Signal

The Fractional Brownian Motion Synthesis Tool lets you save the synthesized signal to disk. The
toolbox creates a MAT-file in the current folder with a name you choose.

To save the synthesized signal from the present selection, use the option File > Save Synthesized
Signal. A dialog box appears that lets you specify a folder and filename for storing the signal. After
saving the signal data to the file fbm07.mat, load the variables into workspace.

load fbm07
whos

Name Size Bytes Class

FBM_PARAMS 1x1 1296 struct array
fbm07 1x1000 8000 double array

FBM_PARAMS

FBM_PARAMS =

struct with fields:

SEED: [1x1 struct]

Wav: 'db10'
Length: 1000
H: 0.7000
Refinement: 6

The synthesized signal is given by fbm07. In addition, the parameters of the generation are given by
FBM_PARAMS, which is a structure array with five fields.

2-44
 Additional Wavelet GUIs

New Wavelet for CWT Using the Pattern Adapted Admissible Wavelet
Design Tool
The toolbox requires only one function to design a new wavelet adapted to a given pattern for CWT:
pat2cwav. You'll find full information about this function in its reference page.

In this section, you'll learn how to

• Load a pattern
• Synthesize a new wavelet adapted to the given pattern
• Detect patterns by CWT using the adapted wavelet
• Compare the detection using both the adapted wavelet and well-known wavelets
• Save the synthesized wavelet

The principle for designing a new wavelet for CWT is to approximate a given pattern using least
squares optimization under constraints leading to an admissible wavelet well suited for the pattern
detection using the continuous wavelet transform (see [MisMOP03] in “References” on page 1-93).

1 Start the New Wavelet for CWT Tool.

From the MATLAB prompt, type nwavtool. The Pattern Adapted Admissible Wavelet Design
Tool appears.

The Wavelet Analyzer appears. Click the New Wavelet for CWT menu item to display the
Pattern Adapted Admissible Wavelet Design Tool.

2 Load the original pattern. In the Pattern Adapted Admissible Wavelet Design tool, from the
File menu, choose Load Pattern. When the Load Pattern dialog box appears, navigate to
matlabroot/toolbox/wavelet/wavelet where matlabroot is the MATLAB root folder. Select
ptpssin1.mat. Click the OK button.

2-45
2 Using Wavelets

The MAT-file defining the pattern can contain more than one variable. In that case, the variable Y
is considered if it exists; otherwise, the first variable is considered.

The selected pattern denoted by F is defined on the interval [0,1] and is of integral 0.1592. It is
not a wavelet, but it is a good candidate because it oscillates like a wavelet.
3 Perform pattern approximation.

Accept the default parameters leading to use a polynomial of degree 3 with constraints of
continuity at the borders 0 and 1, to approximate the pattern F. Click the Approximate button.

After a pause for computation, the tool displays the new wavelet in green superimposed with the
original pattern in red.

2-46
 Additional Wavelet GUIs

The result is not really satisfactory. A solution is to increase the polynomial degree to fit better
the pattern.
4 Using the Polynomial Degree menu, increase the degree by selecting 6. Then click the
Approximate button again.

The result is now of good quality and can be used for pattern detection.
5 Pattern detection using the new wavelet.

Click the Run button.

After a pause for computation, the tool displays the running signal and the pattern detection by
CWT using the adapted wavelet.

2-47
2 Using Wavelets

The running signal is the superimposition of two dilated and translated versions of the pattern F,
namely F((t-20)/8) and F((t-40)/4). The two pairs (position, scale) to be detected are
given by (20,8) and (40,4) and are materialized by dashed lines in the lower right graph of the
contour plot of the CWT. The detection is perfect because the two local maxima of the absolute
values of the continuous wavelet coefficients fit perfectly.
6 Using the Running signal frame, select the Noise check box to add an additive noise to the
previous signal. Click the Run button again.

The quality of the detection is not altered at all.

7 Compare the adapted wavelet and well-known wavelets.

Let us now compare the performance for pattern detection of the adapted wavelet versus well-
known wavelets. Click the Compare button. A new window appears.

2-48
 Additional Wavelet GUIs

This tool displays the pattern detection performed with the adapted wavelet on the left and db1
wavelet (default) on the right. The two positions are perfectly detected in both cases but scales
are slightly underestimated by the db1 wavelet.

The tool allows you to generate various running signals and choose the wavelet to be compared
with the adapted one.

Click the Close button to get back to the main window.

Saving the New Wavelet

The New Wavelet for CWT Tool lets you save the synthesized wavelet. The toolbox creates a MAT-file
in the current folder with a name you choose.

To save the new wavelet from the present selection, use the option File > Save Adapted Wavelet. A
dialog box appears that lets you specify a folder and filename for storing the data. After you save the
wavelet data to the file newwavel.mat, the adapted wavelet is given by X and Y.

Note that the version of the saved wavelet is correctly defined to be used in the CWT algorithm and is
such that its square norm is equal to 1.

See Also
Signal Multiresolution Analyzer | Wavelet Analyzer | Wavelet Signal Denoiser

2-49
 3

Getting Started with Wavelet Analysis

• “Wavelet Families and Properties” on page 3-2

• “Visualizing Wavelets, Wavelet Packets, and Wavelet Filters” on page 3-4
• “Continuous Wavelet Analysis” on page 3-7
• “Continuous Wavelet Transform and Inverse Continuous Wavelet Transform” on page 3-9
• “Discrete Wavelet Analysis” on page 3-13
• “Lifting” on page 3-20
• “Critically Sampled Wavelet Packet Analysis” on page 3-27
3 Getting Started with Wavelet Analysis

Wavelet Families and Properties

This example shows how to find and display information about available wavelets. The Wavelet
Toolbox software contains an extensive selection of the most commonly-used wavelets and orthogonal
and biorthogonal wavelet filters. You also have the ability to add your own filters to the toolbox.

Determine the existing wavelet families. Display the wavelet family names in the command window.

waveletfamilies('f')

Display the names of all available wavelets in each family.

waveletfamilies('a')

You can also use wavemngr to display the available wavelet families.

wavemngr('read')

Use the wavelet family short name to determine what analysis an existing wavelet supports.

The wavelet family short name for the Daubechies extremal-phase wavelets is 'db'.

waveinfo('db')

Determine what analysis the Morlet wavelet supports. The wavelet family short name is 'morl'.

waveinfo('morl')

Use the Wavelet Toolbox Wavelet Display tool to investigate wavelet families. To start the interactive
tool, enter wvdtool at the command line.

Select the db4 wavelet and click Display.

3-2
 Wavelet Families and Properties

See Also

More About
• “Visualizing Wavelets, Wavelet Packets, and Wavelet Filters” on page 3-4

3-3
3 Getting Started with Wavelet Analysis

Visualizing Wavelets, Wavelet Packets, and Wavelet Filters

This example shows how to use wfilters, wavefun, and wpfun to obtain the filters, wavelet, or
wavelet packets corresponding to a particular wavelet family. You can visualize 2-D separable
wavelets with wavefun2.

Obtain the decomposition (analysis) and reconstruction (synthesis) filters for the biorthogonal spline
wavelet filters with 3 vanishing moments in the reconstruction filter and 5 vanishing moments in the
decomposition filter.
[LoD,HiD,LoR,HiR] = wfilters('bior3.5');
subplot(2,2,1)
stem(LoD,'markerfacecolor',[0 0 1]); title('Lowpass Decomposition Filter');
subplot(2,2,2)
stem(LoR,'markerfacecolor',[0 0 1]); title('Lowpass Reconstruction Filter');
subplot(2,2,3)
stem(HiD,'markerfacecolor',[0 0 1]); title('Highpass Decomposition Filter');
subplot(2,2,4)
stem(HiR,'markerfacecolor',[0 0 1]); title('Highpass Reconstruction Filter');

Visualize the real-valued Morlet wavelet. There is no associated scaling function.

figure
[psi,xval] = wavefun('morl');
plot(xval,psi,'linewidth',2)
title('$\psi(x) = e^{-x^2/2} \cos{(5x)}$','Interpreter','latex',...
'fontsize',14);

3-4
 Visualizing Wavelets, Wavelet Packets, and Wavelet Filters

Obtain the first 4 wavelet packets for the Daubechies least-asymmetric wavelet with 4 vanishing
moments, sym4.

[wpws,x] = wpfun('sym4',4,10);
for nn = 1:size(wpws,1)
subplot(3,2,nn)
plot(x,wpws(nn,:))
axis tight
title(['W',num2str(nn-1)]);
end

3-5
3 Getting Started with Wavelet Analysis

See Also
cwtfilterbank | dwtfilterbank | wavefun | wpfun

3-6
 Continuous Wavelet Analysis

Continuous Wavelet Analysis

This example shows how to perform time-frequency analysis using the continuous wavelet transform
(CWT). Continuous wavelet analysis provides a time-scale/time-frequency analysis of signals and
images. The Wavelet Toolbox™ software has both command line and interactive functionality to
support continuous wavelet analysis of 1-D signals.

Construct a signal consisting of two sinusoids with frequencies of 100 and 50 Hz, and white noise.
The support of the two sinusoids is disjoint. The 100-Hz sine wave begins at t = 0 and has a duration
of 1 second. The 100-Hz sinusoid has an amplitude of 2. The 50-Hz sinusoid begins at three seconds
and has a duration of two seconds. The 50-Hz sinusoid has an amplitude of 1. The sampling frequency
is 1 kHz. The signal length is 5000 samples.
Fs = 1000;
t = linspace(0,5,5e3);
x = 2*cos(2*pi*100*t).*(t<1)+cos(2*pi*50*t).*(3<t)+0.3*randn(size(t));

Plot the signal.

plot(t,x)
xlabel('Time (s)')
ylabel('Amplitude')

Use cwt to obtain the CWT of the signal and plot its scalogram. The magnitudes of the sinusoid
components in the colorbar are essentially their amplitudes even though they are at different scales.
cwt(x,Fs)

3-7
3 Getting Started with Wavelet Analysis

See Also
cwt | cwtfilterbank

More About
• “CWT-Based Time-Frequency Analysis”

3-8
 Continuous Wavelet Transform and Inverse Continuous Wavelet Transform

Continuous Wavelet Transform and Inverse Continuous

Wavelet Transform
This example shows how to use the continuous wavelet transform (CWT) and inverse CWT.

CWT of Sine Waves and Impulses

Create and plot a signal consisting of two disjoint sine waves with frequencies of 100 and 50 Hz
punctuated by two impulses. The sampling frequency is 1 kHz and the total signal duration is one
second. The 100-Hz sine wave occurs over the first 250 milliseconds of the data. The 50-Hz sinusoid
occurs over the last 500 milliseconds. The impulses occur at 650 and 750 milliseconds. The signal
also has N(0, 0 . 12) additive white Gaussian noise. The impulse at 650 milliseconds is visible, but the
impulse at 750 milliseconds is not clearly evident in the time-domain data.

Fs = 1000;
t = 0:1/Fs:1-1/Fs;
x = zeros(size(t));
x([625,750]) = 2.5;
x = x+ cos(2*pi*100*t).*(t<0.25)+cos(2*pi*50*t).*(t>=0.5)+...
0.1*randn(size(t));
plot(t.*1000,x)
grid on;
xlabel('msec'); ylabel('Amplitude');

Obtain and plot the CWT using the default analytic Morse wavelet.

3-9
3 Getting Started with Wavelet Analysis

[cfs,f] = cwt(x,1000);
contour(t.*1000,f,abs(cfs));
xlabel('msec'); ylabel('Hz');
grid on;

The CWT moduli correctly show the supports of the disjoint sinusoids and the locations of the
impulses at 650 and 750 milliseconds. In the CWT moduli, the impulse at 750 milliseconds is clearly
visible. This is especially true if you plot just the finest-scale wavelet coefficients.

plot(t.*1000,abs(cfs(1,:)))
grid on
title('Fine-Scale Wavelet Coefficient Moduli')
xlabel('msec')

3-10
 Continuous Wavelet Transform and Inverse Continuous Wavelet Transform

Frequency-Localized Inverse CWT

Using the inverse CWT you can construct frequency-localized approximations to events in your time
series. Use the inverse CWT to obtain an approximation to the 100-Hz sinusoid in the previous
example.

xrec = icwt(cfs,f,[90 110]);

plot(t,x);
hold on;
plot(t,xrec,'r');
legend('Original Signal','Inverse CWT Approximation',...
'Location','NorthEast');
grid on;

3-11
3 Getting Started with Wavelet Analysis

If you zoom in on the plot, you see the 100-Hz component is well approximated but the 50-Hz
component has been removed.

3-12
 Discrete Wavelet Analysis

Discrete Wavelet Analysis

Wavelet Toolbox software enables you to analyze signals, images, and 3-D data using orthogonal and
biorthogonal critically-sampled discrete wavelet analysis. Critically-sampled discrete wavelet analysis
is also known as decimated discrete wavelet analysis. Decimated discrete wavelet analysis is most
appropriate for data compression, denoising, and the sparse representation of certain classes of
signals and images.

In decimated discrete wavelet analysis, the scales and translations are dyadic.

You can perform 1-D, 2-D, and 3-D decimated discrete wavelet analysis using the interactive tool by
entering waveletAnalyzer at the command line and clicking Wavelet 1-D, Wavelet 2-D, or
Wavelet 3-D.

1-D Wavelet Denoising

This example shows how to denoise a signal using discrete wavelet analysis.

Create a reference signal.

len = 2^11;
h = [4 -5 3 -4 5 -4.2 2.1 4.3 -3.1 5.1 -4.2];
t = [0.1 0.13 0.15 0.23 0.25 0.40 0.44 0.65 0.76 0.78 0.81];
h = abs(h);
w = 0.01*[0.5 0.5 0.6 1 1 3 1 1 0.5 0.8 0.5];
tt = linspace(0,1,len);
xref = zeros(1,len);
for j=1:11
xref = xref+(h(j)./(1+((tt-t(j))/w(j)).^4));
end

Add zero-mean white Gaussian noise with a variance of 0.25.

rng default
x = xref + 0.5*randn(size(xref));
plot(x)
axis tight

3-13
3 Getting Started with Wavelet Analysis

Denoise the signal down to level 3 using the Daubechies least asymmetric wavelet with 4 vanishing
moments. Use the universal threshold selection rule of Donoho and Johnstone with soft thresholding
based on the DWT coefficients at level 1. Use the periodization signal extension mode —
dwtmode('per'). Plot the result along with the reference signal for comparison.

origmode = dwtmode('status','nodisplay');
dwtmode('per','nodisplay')
xd = wdenoise(x,3,'Wavelet','sym4',...
'DenoisingMethod','UniversalThreshold','NoiseEstimate','LevelIndependent');
plot(xd)
axis tight
hold on
plot(xref,'r')
legend('Denoised','Reference')

3-14
 Discrete Wavelet Analysis

Restore the original extension mode.

dwtmode(origmode,'nodisplay')

2-D Decimated Discrete Wavelet Analysis

This example shows how to obtain the 2-D DWT of an input image.

Load and display the image. The image consists of vertical, horizontal, and diagonal patterns.

load tartan;
imagesc(X); colormap(gray);

3-15
3 Getting Started with Wavelet Analysis

Obtain the 2-D DWT at level 1 using the biorthogonal B-spline wavelet and scaling filters with 2
vanishing moments in the analysis filters and 4 vanishing moments in the synthesis filters. Extract the
horizontal, vertical, and diagonal wavelet coefficients and the approximation coefficients. Display the
results.

[C,S] = wavedec2(X,1,'bior2.4');
[H,V,D] = detcoef2('all',C,S,1);
A = appcoef2(C,S,'bior2.4');
subplot(221);
imagesc(A); title('Approximation Level 1');
colormap(gray);
subplot(222);
imagesc(H); title('Horizontal Details');
subplot(223);
imagesc(V); title('Vertical Details');
subplot(224);
imagesc(D); title('Diagonal Details');

3-16
 Discrete Wavelet Analysis

You see that the wavelet details are sensitive to particular orientations in the input image. The
approximation coefficients are a lowpass approximation to the original image.

Nondecimated Discrete Wavelet Analysis

This example shows how to obtain the nondecimated (stationary) wavelet transform of a noisy
frequency-modulated signal.

Load the noisy Doppler signal and obtain the stationary wavelet transform down to level 4.
load noisdopp
swc = swt(noisdopp,4,'sym8');

Plot the original signal and the level 1 and 3 wavelet coefficients. Plot the level 4 approximation.
subplot(4,1,1)
plot(noisdopp)
subplot(4,1,2)
plot(swc(1,:))
ylabel('D1')
set(gca,'ytick',[])
subplot(4,1,3)
plot(swc(3,:))
ylabel('D3')
set(gca,'ytick',[])
subplot(4,1,4)

3-17
3 Getting Started with Wavelet Analysis

plot(swc(5,:))
ylabel('A4')
set(gca,'ytick',[])

The wavelet and approximation coefficients at each level are equal in length to the input signal. The
additive noise is almost entirely localized in the level one detail coefficients. The level 3 detail
coefficients capture the high-frequency oscillations at the beginning of the Doppler signal. The level 4
approximation coefficients are a lowpass approximation to the Doppler signal.

Obtain the 2-D nondecimated wavelet transform of an image. Use the Daubechies least asymmetric
wavelet, sym4, and obtain the multiresolution analysis down to level 3. Load the image. Use
wcodemat to scale the matrix for display.

load tartan
nbcol = size(map,1);
cod_X = wcodemat(X,nbcol);

Obtain the nondecimated multiresolution analysis down to level 3.

[ca,chd,cvd,cdd] = swt2(X,3,'sym4');

Display the original image and the approximation and detail coefficients at each level.

figure
subplot(2,2,1)
image(cod_X)
title('Original Image')

3-18
 Discrete Wavelet Analysis

colormap(map)

for k = 1:3
cod_ca = wcodemat(ca(:,:,k),nbcol);
cod_chd = wcodemat(chd(:,:,k),nbcol);
cod_cvd = wcodemat(cvd(:,:,k),nbcol);
cod_cdd = wcodemat(cdd(:,:,k),nbcol);
decl = [cod_ca,cod_chd;cod_cvd,cod_cdd];

subplot(2,2,k+1)
image(decl)

title(['SWT: Approx. ', ...

'and Det. Coefs (Lev. ',num2str(k),')'])
colormap(gray)
end

See Also
dwtfilterbank | modwt | modwtmra | swt | swt2 | wavedec | wavedec2 | wdenoise | wdenoise2

3-19
3 Getting Started with Wavelet Analysis

Lifting
This example shows how to use lifting on a 1-D signal.
2
Create a 1-D signal that is piecewise constant over 2 samples. Add N(0, 0 . 1 ) noise to the signal.
x = [1 1 2 2 -3.5 -3.5 4.3 4.3 6 6 -4.5 -4.5 2.2 2.2 -1.5 -1.5];
x = repmat(x,1,64);
rng default
x = x+ 0.1*randn(size(x));

Plot the signal and zoom in on the first 100 samples to visualize the correlation in neighboring
samples.
plot(x)
xlim([0 100])
title('Signal')
xlabel('Index')
ylabel('Amplitude')

Use the lazy wavelet to obtain the even and odd polyphase components of the signal. Plot the detail
(wavelet) coefficients in D, and observe that this transform has not decorrelated the signal. The
wavelet coefficients look very much like the signal.
LS = liftingScheme;
[A,D] = lwt(x,'LiftingScheme',LS,'Level',1);
plot(D{1})

3-20
 Lifting

xlim([0 100])
title('Detail Coefficients')
xlabel('Index')
ylabel('Amplitude')

Add a prediction lifting step that subtracts the even-indexed coefficient from the odd-coefficient one
sample later, x(2n + 1) − x(2n).

ElemLiftStep = liftingStep('Type','predict','Coefficients',-1,'MaxOrder',0);
LSnew = addlift(LS,ElemLiftStep);

Because the signal is piecewise constant over consecutive samples with additive noise, the new
prediction step should result in wavelet coefficients small in absolute value. In this case, the wavelet
transform does decorrelate the data. Verify this by finding the approximation and detail coefficients
with the new prediction step.

[A,D] = lwt(x,'liftingScheme',LSnew,'Level',1);

If you plot the detail (wavelet) coefficients, you see that the wavelet coefficients no longer resemble
the original signal.

plot(D{1})
xlim([0 100])
title('Detail Coefficients')
xlabel('Index')
ylabel('Amplitude')

3-21
3 Getting Started with Wavelet Analysis

The approximation coefficients, A, of the previous transform constitute the even polyphase
component of the signal. Therefore, the coefficients are affected by aliasing. Use an update lifting
step to update the approximation coefficients and reduce aliasing. The update step replaces the
approximation coefficients by x(2n) + 1/2(x(2n + 1) − x(2n)), which is equal to the average of x(2n)
and x(2n + 1). The averaging is a lowpass filtering, which helps to reduce aliasing.

ElemLiftStep = liftingStep('Type','update','Coefficients',1/2,'MaxOrder',0);
LSnew = addlift(LSnew,ElemLiftStep);

Use the new lifting scheme to obtain the wavelet transform of the input signal. The approximation
coefficients resemble a smooth version of the original signal.

[A,D] = lwt(x,'liftingScheme',LSnew,'Level',1);
plot(A)
xlim([0 100])

3-22
 Lifting

Create a new lifting scheme with the same lifting steps as LSnew. Apply scaling factors to ensure
perfect reconstruction. Obtain the approximation and wavelet coefficients using the lifting scheme
and reconstruct the signal using ilwt. Verify perfect reconstruction.
scaleFactors = [sqrt(2) sqrt(2)/2];
ElemLiftStep1 = liftingStep('Type','predict','Coefficients',-1,'MaxOrder',0);
ElemLiftStep2 = liftingStep('Type','update','Coefficients',1/2,'MaxOrder',0);
LSscale = liftingScheme('LiftingSteps',[ElemLiftStep1;ElemLiftStep2],'NormalizationFactors',scale
[A,D] = lwt(x,'liftingScheme',LSscale,'Level',1);
xrecon = ilwt(A,D,'liftingScheme',LSscale);
max(abs(x(:)-xrecon(:)))

ans = 1.7764e-15

The preceding example designed a wavelet, which effectively removed a zeroth order polynomial
(constant). If the behavior of the signal is better represented by a higher-order polynomial, you can
design a dual wavelet with the appropriate number of vanishing moments to decorrelate the signal.

Use the lifting scheme to design a wavelet with two vanishing moments. A dual wavelet with two
vanishing moments decorrelates a signal with local behavior approximated by a first-order
polynomial. Create a signal characterized by first-order polynomial behavior with additive
N(0, 0 . 252) noise.
y = [1 0 0 4 0 0 -1 0 0 2 0 0 7 0 0 -4 0 0 1 0 0 -3];
x1 = 1:(21/1024):22-(21/1024);
y1 = interp1(1:22,y,x1,'linear');
rng default

3-23
3 Getting Started with Wavelet Analysis

y1 = y1+0.25*randn(size(y1));
plot(x1,y1)
xlim([1 22])

In this case, the wavelet coefficients should remove a first-order polynomial. If the signal value at an
odd index, x(2n + 1), is well approximated by a first-order polynomial fitted to the surrounding sample
values, then 1/2(x(2n) + x(2n + 2)) should provide a good fit for x(2n + 1). In other words, x(2n + 1)
should be the midpoint between x(2n) and x(2n + 2).

It follows that x(2n + 1) − 1/2(x(2n) + x(2n + 2)) should decorrelate the signal.

Create a new lifting scheme with the prediction lifting step which models the preceding equation.

ElemLiftStep = liftingStep('Type','predict','Coefficients',[-1/2 -1/2],'MaxOrder',1);

LS = liftingScheme('LiftingSteps',ElemLiftStep,'NormalizationFactors',1);

Use the lifting scheme to obtain the approximation and detail coefficients and plot the result.

[A,D] = lwt(y1,'LiftingScheme',LS,'Level',1);
subplot(2,1,1)
plot(A)
xlim([1 512])
title('Approximation Coefficients')
subplot(2,1,2)
plot(D{1})
xlim([1 512])
title('Detail Coefficients')

3-24
 Lifting

You see that the wavelet coefficients appear to only contain noise, while the approximation
coefficients represent a denoised version of the original signal. Because the preceding transform uses
only the even polyphase component for the approximation coefficients, you can reduce aliasing by
adding an update step.

Create a new lifting scheme that consists of the previous prediction step and a new update step that
reduces aliasing. Normalize the lifting scheme to produce a perfect reconstruction filter bank. Obtain
the discrete wavelet transform with the new lifting scheme and plot the results.

scaleFactors = [sqrt(2) sqrt(2)/2];

ElemLiftStep1 = liftingStep('Type','predict','Coefficients',[-1/2 -1/2],'MaxOrder',1);
ElemLiftStep2 = liftingStep('Type','update','Coefficients',[1/4 1/4],'MaxOrder',0);
LSnew = liftingScheme('LiftingSteps',[ElemLiftStep1;ElemLiftStep2],'NormalizationFactors',scaleFa
[A,D] = lwt(y1,'liftingScheme',LSnew,'Level',1);
subplot(2,1,1)
plot(A)
xlim([1 512])
title('Approximation Coefficients')
subplot(2,1,2)
plot(D{1})
xlim([1 512])
title('Detail Coefficients')

3-25
3 Getting Started with Wavelet Analysis

Demonstrate that you have designed a perfect reconstruction filter bank.

y2 = ilwt(A,D,'liftingScheme',LSnew);
max(abs(y2(:)-y1(:)))

ans = 1.7764e-15

3-26
 Critically Sampled Wavelet Packet Analysis

Critically Sampled Wavelet Packet Analysis

This example shows how to obtain the wavelet packet transform of a 1-D signal. The example also
demonstrates that frequency ordering is different from Paley ordering.

Create a signal consisting of a sine wave with a frequency of 7π/8 radians/sample in additive white
Gaussian N(0,1/4) noise. The sine wave occurs between samples 128 and 512 of the signal. Set the
dwtmode to periodization and return it to your original setting at the end of the example.

rng default
st = dwtmode('status','nodisplay');
dwtmode('per','nodisp');

n = 0:1023;
indices = (n>127 & n<=512);
x = cos(7*pi/8*n).*indices+0.5*randn(size(n));

Obtain the wavelet packet transform down to level 2 using the Daubechies least asymmetric wavelet
with 4 vanishing moments. Plot the wavelet packet tree.

T = wpdec(x,2,'sym4');
plot(T)

3-27
3 Getting Started with Wavelet Analysis

Find the Paley and frequency ordering of the terminal nodes.

[tn_pal,tn_freq] = otnodes(T);

tn_freq contains the vector [3 4 6 5], which shows that the highest frequency interval, [3π/4, π),
is actually node 5 in the Paley-ordered wavelet packet tree.

Click on node (2,2) in the wavelet packet tree to see that the frequency ordering correctly predicts
the presence of the sine wave.

The wavelet packet transform of a 2-D image yields a quaternary wavelet packet tree. Load an
example image. Use the biorthogonal B-spline wavelet with 3 vanishing moments in the
reconstruction wavelet and 5 vanishing moments in the decomposition wavelet. Plot the resulting
quaternary wavelet packet tree.

load tartan
T = wpdec2(X,2,'bior3.5');
plot(T)

3-28
 Critically Sampled Wavelet Packet Analysis

dwtmode(st,'nodisplay')

3-29