International Journal of Research in Engineering and Science (IJRES

)
ISSN (Online): 2320-9364, ISSN (Print): 2320-9356
www.ijres.org Volume 4 Issue 2 ǁ February. 2016 ǁ PP.31-35

A Short Report on Different Wavelets and Their Structures.
Sumana R. Shesha* and Achala L. Nargund*
*Post Graduate Department of Mathematics and Research Centre in Applied Mathematics MES College,
Malleswaram 15th cross, Bangalore-03

Abstract: This article consists of basics of wavelet analysis required for understanding of and use of wavelet
theory. In this article we briefly discuss about HAAR wavelet transform their space and structures.
Keywords: HAAR wavelet transform, Multiresolution analysis (MRA), Daubechies wavelet transforms,
Biorthogonal wavelets.
Mathematical Classification: 42C40
I. Introduction to wavelets [1,2,6,7]
Wavelet analysis is used to decomposes sounds and images into component waves of varying durations,
called wavelets which are localised vibrations of a sound signal or localized detail in an image. This analysis can
be used in signal processing for removing noise. Jean Baptist Joseph Fourier (1807) has introduced frequency
analysis leading Fourier analysis. He also explained the fact that functions can be represented as the sum of sine
and cosine which led to Fourier Transform.
Alfred Haar in the year 1909 introduced wavelets. Haar’s contribution to wavelets is very evident so the
entire wavelet family named after him. The Haar wavelets are the simplest of the wavelet families.

II. Haar wavelets [2,3,4]
A Haar wavelet is the simplest type of wavelet. In discrete form, Haar wavelets are related to a
mathematical operation called the Haar transform. The Haar transform serves as a prototype for all other
wavelet transforms.
Let  t  be the box function defined by,
1 0  t  1  .
 t   

0 otherwise
The equation,
 t    2t    2t  1

III. Haar spaces
(1)

(2)
is called the dilation equation. The Haar function  t  is typically called a scaling function. The space

V0 generated by the Haar function  t  . The set
. The set  t  k kZ also forms a Riesz basis for V 0 .

V0  span t  k kZ  L2 R  is called the Haar space

 t  k kZ

forms an orthonormal basis for V 0

[Riesz basis: Suppose that  t   L2 R  and 0  A  B are constants such that

A  cn2 
nZ

c
nZ

n

n t   B  cn2
nZ

for any square-summable sequence


c.

(3)
Then we say that the translates

  n t  nZ

form a Riesz basis of

V0  span n t nZ .] The projection of the functions f t   L R  into V0 is given by,
P  f t    f t ,  t  k   t  k 
2

kZ
(4)
The vector space V j  span 2 j t  k kZ  L2 R is called the Haar space V j . V j  L2 R is a vector

space of piecewise constant functions with possible breakpoints at 2  j Z . For each k  Z , define the function
j
 j

(5)
 j ,k  2 2   2 2 t  k  .


The set   j ,k t 
is an orthonormal basis for V j . The compact support for the functions  j , k t  is given
kZ

by,

www.ijres.org

31 | Page

A short report on different Wavelets and their structures.

k  1 .
 k
supp j ,k    j ,
2 j 
2

(6)

The dyadic interval I j , k , j , k  Z , dilation equation and projection function are defined by

k
k  1 .
 k k  1 
I j ,k   j ,
 t  R j  t 

j 
2
2
2
2j 

 
2
2
 j ,k t  
 j 1, 2 k t  
 j 1, 2 k 1 t  .
2
2
P  f t  
 t , f t   t  ,

j

kZ

j ,k

(7)
(8)
(9)

j ,k

where P j is called the approximation operator and the space V j is also known as approximation space. The
Haar spaces  V j 
satisfy
jZ

         V1  V0  V1  V2         

The function f t   V j if and only if f 2t   V j 1 .

V

j

         V1  V0  V1  V2           0

.

(10)

jZ

V

j

          V1  V0  V1  V2           L2 R 

jZ

The wavelet function  t  is given by,

1








 t   2t   2t  1   1

0

IV. Haar wavelet spaces

1
2

1

 t 1 
2
.
otherwise


0t 

(11)

The equation,
 t    2t    2t  1 , is called the dilation equation. The space W0  span t  k kZ  L2 R  is
called the Haar wavelet space W0 generated by the Haar wavelet function t  . The set  t  k kZ forms
an orthonormal basis for W0 . The set  t  k kZ also forms a Riesz basis for W0 . The Haar wavelet space

W0 satisfies the following properties.

If f t   V0 and g t   W0 , then

and W0 are perpendicular to each other.
V1  V0  W0 .

f t , g t   0 .

(12)

V0

The vector space W j  span 2 j t  k kZ  L2 R is called the Haar wavelet space W j .
For each k  Z , define the function

j

 j ,k  2 2  2 j / 2 t  k  . The set

basis for W j . The compact support for the functions 
supp

j ,k



j, k

t  is given by,

k  1 .
 k
,
j

2j 
2

For j, k  Z , the dilation equation for 

 t 
j ,k

kZ

is an orthonormal

(13)

j, k

t  is given by,

www.ijres.org

32 | Page

A short report on different Wavelets and their structures.

j ,k

t  

2
2
 j 1, 2 k t  
 j 1, 2 k 1 t 
2
2

.

(14)

For each j  Z , the detail operator Q j on functions f t   L2 R  is defined by,
Q j  f t   Pj 1  f t   Pj  f t  .

The Haar wavelet spaces

W 

jZ

j

(15)

satisfy the following properties.

If j and l are integers with l  j , then V j and W j are perpendicular to each other.
If j and l are integers with l  j , then W j and W l are perpendicular to each other.
V j 1  V j  W j

.

Let f j 1 t   V j 1 be defined as f t  
j 1

 a m j ,m t  and suppose that h is the vector given by

mZ

T


 2
2  , then the projection of f t  into V can be written as
T
h  h0 , h1   
,
j 1
j

2 
 2

f j t    bk  j ,k t    f j 1 t ,  j ,k t   j ,k t 
kZ

where b 
k

(16)

kZ

k

2
a 2 k  a 2 k 1   h  a k for k  Z and a  a2k , a2k 1
2

T .

Let f j 1 t   V j 1 be defined as
f j 1 t  

a

mZ

m

 j 1, m t  .

(17)

Suppose that f j t  is the projection of f j 1 t  into V j and g is the vector given by
T

 2

2 .
T
g  g 0 , g 1   
,

2 
 2

(18)

If g j t   f j 1 t   f j t  is the residual function in V j 1 , then g j t   W j and is thus given by
g j t  

 c  t 
kZ

where c k 

k

(19)

j ,k

2
a 2 k  a 2 k 1   g  a k for
2

k Z


and a k  a2k , a2k 1 T .

Thus, a piecewise constant function f j t , j  0 with possible breakpoints at points in 2 j Z can be
decomposed into a piecewise constant approximation function f j 1 t  with possible breakpoints on the coarser
grid 2 j 1 Z and a piecewise constant detail function g j 1 t  with possible breakpoints in 2 j Z . This process
can be iterated and finally f j t  can be written as the sum of an approximation function f 0 t  with possible
breakpoints at the integers and detail functions g 0 t , g 1 t ,      , g j 1 t  with possible breakpoints at
1
1
Z,
Z ,  , 2 j Z
2
4

respectively. The next step is to model the decomposition in terms of linear

transformations (matrices). The technique to process discrete data via a discrete version of the decomposition
process is called the discrete Haar wavelet transformation.

V. Discrete Haar wavelet transformation
www.ijres.org

33 | Page

A short report on different Wavelets and their structures.

WN

The

Suppose that N is an even positive integer. The discrete Haar wavelet transformation is defined as,


2
2
0
0

0
0


2
2


2
2


0
0

0
0


2
2
        

2
2 
(20)

0
0
0
0

HN 2  
2
2





 G N 2   2  2
0
0

0
0 
2
 2

2
2


0
0






0
0


2
2
                                   

2
2 
 0
0
0
0


2
2 

N
N
2

block H N

2

is called the averages block and the N  N block G N
2

2

is called the details block.





We have, H N 2 a
 b and G N 2 a  c . The inverse discrete Haar wavelet transformation is given by,

b  .

T
a  WN   
c 

(21)

The two-dimensional discrete Haar wavelet transformation B of a matrix A is defines as,
B  WM A W NT .

The inverse of two-dimensional discrete Haar wavelet transformation is given by,
A  W MT B W N .

VI. Advantages of Haar Wavelet Transform
a)
b)
c)
d)

It is very simple.
It is fast.
It can be calculated without any need of temporary array so it needs less memory.
It can be reversed exactly without the edge effects.

Original function

f  sin x

VII. GRAPHS [5]

www.ijres.org

34 | Page

A short report on different Wavelets and their structures.
Haar Wavelet

Haar Wavelet for

f  sin x

VIII.

Acknowledgement

We are very much thankful to Prof. N. M. Bujurke, UGC Emeritus professor and INSA Senior Scientist for
introducing us to such a deep subject and helping us in preparing this report. The first author would like to thank
the three academies, namely, the Indian Academy of Sciences, Bangalore, the Indian National Science
Academy, New Delhi and the National Academy of Sciences India, Allahabad for providing an opportunity to
study this subject under Prof. N. M. Bujurke through Summer Research Fellowship Programme.

Refernces
[1.]
[2.]
[3.]
[4.]
[5.]
[6.]
[7.]

A brief history of Wavelets – Brendon Keift - Internet
A Primer on WAVELETS and their Scientific Applications – James S. Walker –Chapman &
Hall/CRC.
WAVELET THEORY: An Elementary Approach with Applications – David K. Ruch & Patrick J. Van
Fleet – Wiley.
An Introduction to Wavelet Analysis – David F. Walnut – Birkhäuser.
Wavelets and Multiwavelets – Fritz Keinert – Chapman & Hall/CRC.
Wavelets: Tools for Science and Technology – Stephane Jaffard, Yves Meyer, Robert D. Ryan – Siam.
THE WORLD According to WAVELETS: The Story of a Mathematical Technique in the Making –
Barbara Burke Hubbard – Universities Press.

www.ijres.org

35 | Page