The discrete wavelet transform (DWT) is a
representation of a signal
xt∈
ℒ
2
x
t
ℒ
2
using an orthonormal basis consisting of a
countably-infinite set of wavelets. Denoting the
wavelet basis as
ψ
k
,
n
t
k∈Z∧n∈Z
ψ
k
,
n
t
k
n
, the DWT transform pair is

xt=∑k=−∞∞∑n=−∞∞
d
k
,
n
ψ
k
,
n
t
x
t
k
n
d
k
,
n
ψ
k
,
n
t

(1)
d
k
,
n
=
ψ
k
,
n
t · xt=∫−∞∞
ψ
k
,
n
t*xtdt
d
k
,
n
ψ
k
,
n
t
x
t
t
ψ
k
,
n
t
x
t

(2)
where

d
k
,
n
d
k
,
n
are the wavelet coefficients. Note the relationship
to Fourier series and to the sampling theorem: in both cases
we can perfectly describe a continuous-time signal

xt
x
t
using a countably-infinite (

i.e.,
discrete) set of coefficients. Specifically, Fourier series
enabled us to describe

periodic signals using
Fourier coefficients

Xk
k∈Z
X
k
k
, while the sampling theorem enabled us to describe

bandlimited signals using signal samples

xn
n∈Z
x
n
n
. In both cases, signals within a limited class are
represented using a coefficient set with a single countable
index. The DWT can describe

*any* signal
in

ℒ
2
ℒ
2
using a coefficient set parameterized by two countable
indices:

d
k
,
n
k∈Z∧n∈Z
d
k
,
n
k
n
.

Wavelets are orthonormal functions in
ℒ
2
ℒ
2
obtained by shifting and stretching a mother
wavelet,
ψt∈
ℒ
2
ψ
t
ℒ
2
. For example,

ψ
k
,
n
t=2−k2ψ2−kt−n ,
k∧n∈Z
k
n
k
n
ψ
k
,
n
t
2
k
2
ψ
2
k
t
n

(3)
defines a family of wavelets

ψ
k
,
n
t
k∈Z∧n∈Z
ψ
k
,
n
t
k
n
related by power-of-two stretches. As

kk increases, the wavelet
stretches by a factor of two; as

nn increases, the wavelet shifts
right.

When
∥ψt∥=1
ψ
t
1
, the normalization ensures that
∥
ψ
k
,
n
t∥=1
ψ
k
,
n
t
1
for all
k∈Z
k
,
n∈Z
n
.

Power-of-two stretching is a convenient, though somewhat
arbitrary, choice. In our treatment of the discrete wavelet
transform, however, we will focus on this choice. Even with
power-of two stretches, there are various possibilities for

ψt
ψ
t
, each giving a different flavor of DWT.

Wavelets are constructed so that
ψ
k
,
n
t
n∈Z
ψ
k
,
n
t
n
(i.e., the set of all shifted
wavelets at fixed scale kk),
describes a particular level of 'detail' in the signal. As
kk becomes smaller
(i.e., closer to
−∞
), the wavelets become more "fine grained" and the
level of detail increases. In this way, the DWT can give a
multi-resolution description of a signal, very
useful in analyzing "real-world" signals. Essentially, the
DWT gives us a *discrete multi-resolution description
of a continuous-time signal in*
ℒ
2
ℒ
2
.

In the modules that follow, these DWT concepts will be
developed "from scratch" using Hilbert space principles. To
aid the development, we make use of the so-called
scaling function
φt∈
ℒ
2
φ
t
ℒ
2
, which will be used to approximate the signal
*up to a particular level of detail*. Like
with wavelets, a family of scaling functions can be
constructed via shifts and power-of-two stretches

φ
k
,
n
t=2−k2φ2−kt−n ,
k∧n∈Z
k
n
k
n
φ
k
,
n
t
2
k
2
φ
2
k
t
n

(4)
given mother scaling function

φt
φ
t
. The relationships between wavelets and scaling
functions will be elaborated upon later via

theory and

example.

The inner-product expression for

d
k
,
n
d
k
,
n
,

Equation 2 is written for the general complex-valued
case. In our treatment of the discrete wavelet transform,
however, we will assume real-valued signals and wavelets.
For this reason, we omit the complex conjugations in the
remainder of our DWT discussions