See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341650998

Unitary Extension Principle for Nonuniform Wavelet Frames in L^2 (R)

Article  in  Journal of Mathematical Physics, Analysis, Geometry · May 2021

DOI: 10.15407/mag17.01.079

CITATIONS READS

2 263

2 authors:

Hari Krishan Malhotra Lalit Kumar Vashisht

University of Delhi University of Delhi
5 PUBLICATIONS   13 CITATIONS    57 PUBLICATIONS   343 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Unitary extension principal View project

All content following this page was uploaded by Lalit Kumar Vashisht on 18 May 2021.

The user has requested enhancement of the downloaded file.

Journal of Mathematical Physics, Analysis, Geometry
2021, Vol. 17, No. 1, pp. 79–94
doi:

Unitary Extension Principle for Nonuniform

Wavelet Frames in L2(R)
Hari Krishan Malhotra and Lalit Kumar Vashisht
Parseval frames have attracted attention of engineers and physicists due
to their potential applications in signal processing. In this paper, we study
the construction of nonuniform Parseval wavelet frames for the Lebesgue
space L2 (R), where the related translation set is not necessary a group. The
main purpose of this paper is to prove the unitary extension principle (UEP)
and the oblique extension principle (OEP) for the construction of multi-
generated nonuniform Parseval wavelet frames for L2 (R). Some examples
are also given to illustrate the results.
Key words: Hilbert frame, nonuniform wavelet system, unitary extension
principle
Mathematical Subject Classification 2010: 42C40; 42C15; 42C30; 42C05

1. Introduction
Wavelets have been extensively studied over last decades and their role in
both pure and applied mathematics is well known. As it is not possible to give a
complete list of applications of wavelets, we at least mention some [1,2,7–9,16,18,
19, 21, 25], see also references therein. Wavelets in L2 (R) are a very efficient tool
as they give orthonormal basis for L2 (R) in the form of dilation and translation
of a finite number of functions in L2 (R), which is a very simple and convenient
form of basis for L2 (R). Gabardo and Nashed [14] considered a generalization of
Mallat’s classic multiresolution analysis (MRA), which is based on the theory of
spectral pairs.
Definition 1.1 ([14, Definition 3.1]). Let N ≥ 1 be a positive integer and
r be an odd integer relatively prime to N such that 1 ≤ r ≤ 2N − 1, an associ-
ated nonuniform multiresolution analysis (abbreviated NUMRA) is a collection
{Vj }j∈Z of closed subspaces of L2 (R) satisfying the following properties:
(i) Vj ⊂ Vj+1 for all j ∈ Z,
2
S
(ii) j∈Z Vj is dense in L (R),
T
(iii) j∈Z Vj = {0},

(iv) f (x) ∈ Vj if and only if f (2N x) ∈ Vj+1 ,

© Hari Krishan Malhotra and Lalit Kumar Vashisht, 2021

80 Hari Krishan Malhotra and Lalit Kumar Vashisht

(v) there exists a function φ ∈ V0 , called the scaling function, such that the
collection {φ(x−λ)}λ∈Λ , where Λ = {0, r/N }+2Z, is a complete orthonormal
system for V0 .

Here, the translate set Λ = {0, r/N }+2Z may not be a group. One can observe
that the standard definition of a one-dimensional multiresolution analysis with
dilation factor equal to 2 is a special case of NUMRA given in Definition 1.1.
Gabardo and Yu [15] considered the sets of nonuniform wavelets in L2 (R) related
to one-dimensional spectral pairs. For fundamental properties of nonuniform
wavelets based on spectral pairs, we refer to [14, 15, 23].
Ron and Shen [20] introduced the unitary extension principle for constructing
a multi-generated tight wavelet frame for L2 (Rd ) based on a given refinable func-
tion. Tight wavelet frames give a more convenient way to represent a function
in L2 (R) in comparison with non-tight wavelet frames, as in that case the frame
operator is a constant multiple of the identity operator in L2 (R). Christensen
and Goh in [6] generalized the unitary extension principle to the locally com-
pact abelian groups. They gave general constructions, based on B-splines on the
group itself as well as on the characteristic functions on the dual group. Moti-
vated by the work of Gabardo and Nashed [14] for the construction of nonuniform
wavelets and application of frames in applied and pure mathematics, we study
nonuniform wavelet frames for the Lebesgue space L2 (R). A notable contribution
of the paper is to introduce the unitary extension principle for the construction
of multi-generated tight nonuniform wavelet frames of the form
j j
{Ψj,λ,` } j∈Z,λ∈Λ = {(2N ) 2 ψ1 ((2N )j γ −λ)} j∈Z ∪· · ·∪{(2N ) 2 ψn ((2N )j γ −λ)} j∈Z
`=1,2,..., n λ∈Λ λ∈Λ

in L2 (R).

1.1. Overview and main results. The paper is organized as follows. In

Section 2, we give basic notations, definitions and properties of operators related
with nonuniform wavelet frames in L2 (R). The general setup for the nonuniform
wavelet frame system in L2 (R) is given in Section 3. Section 4 gives some aux-
iliary results needed in the rest of the paper. The main results are contained
in Section 5. Theorem 5.1 gives the unitary extension principle (UEP) for the
construction of multi-generated tight nonuniform wavelet frames for L2 (R). The
extended version of UEP (or oblique extension principle) for nonuniform wavelet
frames for L2 (R) can be found in Theorem 5.2. Some examples are given in
Section 6 to illustrate our results.

1.2. Relation to the existing work and motivation Duffin and Scha-
effer [13] introduced the concept of a frame for separable Hilbert spaces, while
addressing some difficult problems from the theory of nonharmonic analysis. Let
H be an infinite-dimensional separable Hilbert space with inner p product h·, ·i.
The norm induced by the inner product h·, ·i is given by kf k = hf, f i, f ∈ H.
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 81

A family {fk }∞
k=1 ⊂ H is called a frame for H if there exist positive scalars Ao ≤
Bo < ∞ such that for all f ∈ H,
∞
X
2
Ao kf k ≤ |hf, fk i|2 ≤ Bo kf k2 . (1.1)
k=1

The scalars Ao and Bo are called the lower frame bound and the upper frame
bound, respectively. If it is possible to choose Ao = Bo , then we say that {fk }∞
k=1
is a Ao -Parseval frame (or Ao -tight frame); and a Parseval frame if Ao = Bo = 1.
If only the upper inequality in (1.1) holds, then we say that {fk }∞ k=1 is a Bessel
sequence with Bessel ∞
bound Bo . If {fk }k=1 is a frame for H, then S : H → H,
given by Sf = ∞
P
k=1 hf, fk ifk , is a frame operator which is bounded linear and
invertible on H. This gives the reconstruction formula of each member f ∈ H,
∞
X
f = SS −1 f = hS −1 f, fk ifk .
k=1

Thus, each member of H has an explicit series expansion which need not be
unique. For application of frames in both pure and applied mathematics, we refer
to books of Casazza and Kutyniok [3], Christensen [5], Han [16], Heil [17] and
Krivoshein, Protasov and Skopina [19]. Nowadays, the theory of iterated function
systems, quantum mechanics and wavelets are emerging in important applications
in the frame theory, see [12, 22, 24]. A very recent work on discrete frames of
translates and discrete wavelet frames and their duals in finite dimensional spaces
can be found in [10, 11]. Wavelet frames in L2 (R) are also a very powerful tool
for representing functions in L2 (R) as a sum of series of functions which are the
dilation and translation of a finite number of functions in L2 (R). It provides us
with a convenient tool to expand functions in L2 (R) of a similar type that arise
in orthonormal basis, however, the wavelet frame conditions are weaker which
makes wavelet frames more flexible. Nonuniform wavelet frames could be used
in signal processing, sampling theory, speech recognition and various other areas,
where instead of integer shifts nonuniform shifts are needed.
Motivated by the work of Gabardo and Nashed [14] and Gabardo and Yu [15],
we study the frame properties of nonuniform wavelets in the Lebesgue space
L2 (R). We recall that the extension problems in the frame theory have a long
history. It is showed in [4] that the extension problem has a solution in the
sense that “any Bessel sequence can be extended to a tight frame by adjoining
a suitable family of vectors in the underlying space.” Ron and Shen introduced
the unitary extension principle for the construction of tight wavelet frames in
the Lebesgue space L2 (Rd ). The unitary extension principle allows the construc-
tion of tight wavelet frames with compact support of a desired smoothness and
a good approximation of functions. In real-life applications, all signals are not
obtained from uniform shifts. So there is a natural question regarding analysis
and decompositions of these types of signals with a stable mathematical tool.
Gabardo and Nashed [14] and Gabardo and Yu [15] filled this gap by the concept
of nonuniform multiresolution analysis. In the direction of construction of Par-
seval frames from nonuniform multiwavelet systems, we develop a general setup
82 Hari Krishan Malhotra and Lalit Kumar Vashisht

and prove the unitary extension principle for the construction of multi-generated
nonuniform tight wavelet frames for L2 (R). Ron and Shen [20] gave the unitary
extension principle, where the conditions for the construction of multi-generated
tight wavelet frames for the Lebesgue space L2 (Rd ) are based on a given refinable
function.

2. Preliminaries
As is standard, Z, N and R denote the set of all integers, positive integers
and real numbers, respectively. Throughout the paper, N ∈ N, r being an odd
integer relative prime to N such that 1 ≤ r ≤ 2N − 1 and Λ = {0, r/N } + 2Z.
Notice that the discrete set Λ is not always a group. The support of a function
ψ is denoted by Supp ψ and defined as

Supp ψ := closure of the set {x : ψ(x) 6= 0}.

The set of all continuous functions defined on R with compact support is denoted
by Cc (R). The symbol z denotes the complex conjugate of a complex number z.
The conjugate transpose of a matrix H is denoted by H ∗ , and the bold number
1 denotes the identity matrix. The characteristic function of a set E is denoted
by χE . The spaces L2 (R) and L∞ (R) denote the equivalence classes of square-
integrable functions and essentially bounded functions on R, respectively. Next,
we recall the Parseval identity. Let {ek }k∈Z be an orthonormal basis for a Hilbert
space H. Then, for every f in H, we have
X
|hf, ek i|2 = kf k2 (Parseval identity).
k∈Z

For a, b ∈ R, we consider the following operators on L2 (R):

Ta : L2 (R) → L2 (R), Ta f (γ) = f (γ − a) (Translation by a),

Eb : L2 (R) → L2 (R), Eb f (γ) = e2πibγ f (γ) (Modulation by b),
2 2
√
L : L (R) → L (R), Lf (γ) = 2N f (2N γ) (N-Dilation operator).

The j fold N -dilation, where j ∈ Z, is given by

j
Lj f (γ) = (2N ) 2 f ((2N )j γ).

Definition 2.1. Let {ψ1 , ψ2 , . . . , ψn } ⊂ L2 (R) be a finite set. The family

j j
{Lj Tλ ψ` } j∈Z,λ∈Λ = {(2N ) 2 ψ1 ((2N )j γ−λ)} j∈Z ∪· · ·∪{(2N ) 2 ψn ((2N )j γ−λ)} j∈Z
`=1,2,...,n λ∈Λ λ∈Λ

is called a nonuniform wavelet frame for L2 (R) if there exist finite positive con-
stants A and B such that
n
XXX
Akf k2 ≤ |hf, Lj Tλ ψ` i|2 ≤ Bkf k2 for all f ∈ L2 (R).
j∈Z λ∈Λ `=1
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 83

The Fourier transform of a function f ∈ L1 (R)∩L2 (R) is denoted by Ff or fb

and defined as
Z ∞
Ff = f (γ) =
b f (x)e−2πixγ dx.
−∞

A function f is said to be bandlimited if its Fourier transform fb has a compact

support.
For N ∈ N, j ∈ Z and a ∈ R, by direct calculation, we have the following
properties:
(i) Lj : L2 (R) → L2 (R) is an unitary map,

(ii) Lj Ta = T(2N )−j a Lj ,

(iii) FLj = L−j F,

(iv) FTa = E−a F.

The following lemma shows that it is enough to check the Besselness and frame
condition on a dense subset of the underlying Hilbert space H.

Lemma 2.2 ([17]). Let {fk }k∈I , where I is a countable set, be a sequence of
elements in a Hilbert space H.
(i) If there exists a constant B > 0 such that
X
|hf, fk i|2 ≤ B kf k2
k∈I

for all f in a dense subset D of H, then {fk }k∈I is a Bessel sequence with
bound B.

(ii) If there exist constants A, B > 0 such that

X
A kf k2 ≤ |hf, fk i|2 ≤ B kf k2
k∈I

for all f in a dense subset D of H, then {fk }k∈I is a frame for H with bound B.

3. The nonuniform general setup

In this section, we give a list of assumptions which will be used in the con-
struction of Parseval nonuniform wavelet frames. To be precise, in formulation of
the unitary extension principle there is a long list of assumptions, so instead of
writing each assumption again and again, we state all assumptions and call them
nonuniform general setup: Let ψ0 ∈ L2 (R) be such that
(i) ψb0 (2N γ) = H0 (γ)ψb0 (γ), H0 (γ) ∈ L∞ (R);

(ii) Supp ψb0 (γ) ⊆ [0, 1/(4N )]; and

84 Hari Krishan Malhotra and Lalit Kumar Vashisht

(iii) lim ψb0 (γ) = 1.

γ→0+

Further, let H1 , H2 , . . . , Hn ∈ L∞ (R), and define ψ1 , ψ2 , . . . , ψn ∈ L2 (R) such

that

ψb` (2N γ) = H` (γ)ψb0 (γ), ` = 1, 2, . . . , n.

Let H(γ) be an (n + 1) × 1 matrix given by

 
H0 (γ)
 H1 (γ) 
H(γ) =  .  .
 
 .. 
Hn (γ) (n+1)×1

Then the collection {ψ` , H` }n`=0 is called a nonuniform general setup.

4. Some auxiliary results

In this section, we give some auxiliary results that will be used in the sequel.
Lemma 4.1. Assume that
(i) ψ0 ∈ L2 (R), lim ψb0 (γ) = 1 and Supp ψb0 (γ) ⊆ [0, 1/2];
γ→0+

(ii) f ∈ L2 (R) such that fb ∈ Cc (R).

Then, for any  > 0, there exist J ∈ Z such that
X
(1 − )kf k2 ≤ |hf, Lj Tλ ψ0 i|2 ≤ (1 + )kf k2 for all j ≥ J.
λ∈Λ

Proof. Using Supp ψb0 (γ) ⊆ [0, 1/2] √ and the Parseval identity on L2 (0, 1/2)
with respect to the orthonormal basis { 2e2πi(2m)γ }m∈Z , we compute
X X D E2
2
|hf, Tλ ψ0 i| =  fb, Td
λψ0 

λ∈Λ λ∈Λ
 2
X Z 12 
2πi(2m)γ
= f (γ)ψ0 (γ)e dγ 
b b 

 0 
m∈Z
 2
X Z 12 r

2πi( +2m)γ
+ f (γ)ψ0 (γ)e dγ 
b b N


 0 
m∈Z
Z 1 2
2 
= fb(γ)ψb0 (γ) dγ. (4.1)

0

Under the assumption ψb0 (γ) → 1 as γ → 0+ , it follows that for any  > 0 there
exists a sufficiently small positive real number b = b() such that
X
(1 − )kfbk2 ≤ |hf, Tλ ψ0 i|2 ≤ (1 + )kfbk2 ,
λ∈Λ
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 85

whenever Supp fb ⊆ [−b, b]. Now, in order to prove our result, it remains to
note that for every bandlimited function f , we have hf, Lj Tλ ψ0 i = hL−j f, Tλ ψ0 i,
kL−j f k = kf k, and Supp L[
−j f is supported in [−b, b] for large enough j. This

concludes the proof.

Lemma 4.2. Suppose that

(i) ψ0 ∈ L2 (R) satisfies Supp ψb0 ⊆ [0, 1/(4N )] and ψb0 (2N γ) = H0 (γ)ψb0 (γ),
where H0 (γ) ∈ L∞ (R);

(ii) f ∈ L2 (R) with fb ∈ Cc (R), and H1 , H2 , . . . , Hn ∈ L∞ (R) such that the

(n + 1) × 1 matrix  
H0 (γ)
 H1 (γ) 
H(γ) =  . 
 
 .. 
Hn (γ) (n+1)×1
∗
satisfies H(γ) H(γ) = 1 a.e.;

(iii) ψ1 , ψ2 , . . . ψn ∈ L2 (R) such that ψb` (2N γ) = H` (γ)ψb0 (γ), ` = 1, 2, . . . n.

Then
n X
X X
|hf, Lj−1 Tλ ψ` i|2 = |hf, Lj Tλ ψ0 i|2 .
`=0 λ∈Λ λ∈Λ

Proof. For any j ∈ Z and for any ` = 0, 1, . . . n, we have

hf, Lj−1 Tλ ψ` i = hL−j f, L−1 Tλ ψ` i = hL−j f, T(2N )λ L−1 ψ` i = hLj fb, E−(2N )λ Lψb` i
Z √
= (Lj fb)(γ) 2N ψb` (2N γ)e2πi(2N λ)γ dγ
R
√ Z
= 2N (Lj fb)(γ)H` (γ)ψb0 (γ)e2πi(2N λ)γ dγ. (4.2)
R

Using Supp ψ c0 ⊆ [0, 1/(4N )], and the Parseval identity on L2 (0, 1/(4N )) with
√
respect to the orthonormal basis {2 N e2πi(4N m)γ }m∈Z , we have
X X X
|hf, Lj−1 Tλ ψ` i|2 = |hf, Lj−1 Tλ ψ` i|2 + |hf, Lj−1 Tλ ψ` i|2
λ∈Λ λ∈2Z r
λ∈( N +2Z)
Z 1 2
1 X  4N j b 2πi(4N m)γ
√ 
= (L f )(γ)H` (γ)ψ0 (γ)e 2 N dγ 
b 
2

 0 
m∈Z
Z 1 2
1 X  4N j b 2πi(2r)γ 2πi(4N m)γ
√ 
+ (L f )(γ)H` (γ)ψ0 (γ)e e 2 N dγ 
b 
2

 0 
m∈Z
Z 1 2
1 4N  j b
= (L f )(γ)H` (γ)ψb0 (γ) dγ

2 0
86 Hari Krishan Malhotra and Lalit Kumar Vashisht

Z 1 2
1 4N

 jb
+ (L f )(γ)H` (γ)ψb0 (γ) dγ

2 0
Z 1  2
4N  jb
= (L f )(γ)H` (γ)ψ0 (γ) dγ.
b 
0

Since H(γ)∗ H(γ) = 1 a.e., we have

n X Z 1  2
X 4N
|hf, Lj−1 Tλ ψ` i|2 =
 jb
(L f )(γ)ψb0 (γ) dγ. (4.3)

`=0 λ∈Λ 0

Also,
X X X
|hf, Lj Tλ ψ0 i|2 = |hf, Lj Tλ ψ0 i|2 + |hf, Lj Tλ ψ0 i|2
λ∈Λ λ∈2Z r
λ∈( N +2Z)
 2
X Z
jb 2πi(2m)γ

=  (L f )(γ)ψ0 (γ)e
 b dγ 
m∈Z R
 2
X Z r
j 2πi( +2m)γ

+  (L fb)(γ)ψb0 (γ)e
 N dγ  . (4.4)
m∈Z R

Using Supp ψb0 ⊆ [0, 1/(4N )] ⊂ [0, 1/2] and applying the Parseval formula on
√ 2πi(2m)γ
2
L (0, 1/2) with respect to the orthonormal basis { 2e }m∈Z , we compute
 2
X 1 X Z 12 √ 
|hf, Lj Tλ ψ0 i|2 = (Lj fb)(γ)ψb0 (γ) 2e2πi(2m)γ dγ 

2

m∈Z 0
 
λ∈Λ
Z 1 2
1 X  2 j b √ 2πi( r +2m)γ 
+ (L f )(γ)ψb0 (γ) 2e N dγ 
2

 0 
m∈Z
Z 1 2 Z 1 2
1 2  j b 1 2  j b
= (L f )(γ) ψ (γ) dγ + (L f )(γ) ψ (γ)  dγ
 
0 0
b b
2 0 2 0
  
Z 1 2 Z 1  2
2  4N 
jb jb
= (L f )(γ)ψ0 (γ) dγ = (L f )(γ)ψ0 (γ) dγ. (4.5)
b  b 
0 0

The proof now follows from (4.3) and (4.5).

Lemma 4.3. Let {ψ` , H` }n`=0 be a nonuniform general setup, and let
H(γ)∗ H(γ) = 1. Then the following holds:
(i) {Tλ ψ0 }λ∈Λ is the Bessel sequence with Bessel bound 1.

(ii) For any f ∈ L2 (R),

X
lim |hf, Lj Tλ ψ0 i|2 = 0.
j→−∞
λ∈Λ
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 87

Proof. (i) Let f ∈ L2 (R) be such that fb ∈ Cc (R), and let  > 0 be given.
Then, by Lemma 4.1, we can find an integer j > 0 such that
X
|hf, Lj Tλ ψ0 i|2 ≤ (1 + )kf k2 . (4.6)
λ∈Λ

Also, by Lemma 4.2, we have

X X
|hf, Lj−1 Tλ ψ` i|2 ≤ |hf, Lj Tλ ψ0 i|2 . (4.7)
λ∈Λ λ∈Λ

Applying (4.7) j times and using (4.6), we get

X X
|hf, Tλ ψ0 i|2 ≤ |hf, Lj Tλ ψ0 i|2 ≤ (1 + )kf k2 .
λ∈Λ λ∈Λ

Since  > 0 is arbitrary, we have

X
|hf, Tλ ψ0 i|2 ≤ kf k2 .
λ∈Λ

Because this inequality holds on a dense subset of L2 (R), therefore, by Lemma

2.2, it holds on L2 (R). This proves (i).
(ii) Let f ∈ L2 (R). Since Lj is an unitary map for all j ∈ Z, by using (i), the
family {Lj Tλ ψ0 }λ∈Λ is the Bessel sequence with Bessel bound 1. For any j ∈ Z
and for any bounded interval I ⊂ R, we have
X X X
|hf, Lj Tλ ψ0 i|2 ≤ 2 |hf χI , Lj Tλ ψ0 i|2 + 2 |hf (1 − χI ), Lj Tλ ψ0 i|2
λ∈Λ λ∈Λ λ∈Λ
X
≤2 |hf χI , L Tλ ψ0 i| + 2kf (1 − χI )k2 .
j 2

λ∈Λ

Now, kf (1 − χI )k2 → 0 if we choose I to be sufficiently large. Therefore we only

need to show
X
|hf χI , Lj Tλ ψ0 i|2 → 0 as j → −∞.
λ∈Λ

Using the Cauchy–Schwarz inequality for integrals, we obtain

X X Z 2
j 2 j

|hf χI , L Tλ ψ0 i| = (2N ) j
 f (γ)ψ0 ((2N ) γ − λ) dγ 
 
λ∈Λ λ∈Λ I
XZ 
j 2 ψ0 ((2N )j γ − λ)2 dγ

≤ (2N ) kf k
λ∈Λ I
XZ
= kf k2 |ψ0 (γ)|2 dγ. (4.8)
λ∈Λ (2N )j I−λ

Applying the Lebesgue dominated convergence theorem in (4.8), we have

X
|hf χI , Lj Tλ ψ0 i|2 → 0 as j → −∞.
λ∈Λ

Thus (ii) is proved.

88 Hari Krishan Malhotra and Lalit Kumar Vashisht

5. The unitary extension principle for nonuniform wavelet

frames
We begin this section with the UEP for nonuniform wavelet frames for L2 (R).
Theorem 5.1. Let {ψ` , H` }n`=0 be a nonuniform general setup and
H(γ)∗ H(γ) = 1. Then the nonuniform multiwavelet system {Lj Tλ ψ` } j∈Z,λ∈Λ
`=1,2,...,n
constitutes a Parseval frame for L2 (R).
Proof. Let  > 0 be given. Consider a function f ∈ L2 (R) such that fb ∈
Cc (R). By Lemma 4.1, we can choose J > 0 such that for all j ≥ J,
X
(1 − )kf k2 ≤ |hf, Lj Tλ ψ0 i|2 ≤ (1 + )kf k2 . (5.1)
λ∈Λ

Using Lemma 4.2, we have

X n X
X
|hf, Lj Tλ ψ0 i|2 = |hf, Lj−1 Tλ ψ` i|2
λ∈Λ `=0 λ∈Λ
X n X
X
j−1 2
= |hf, L Tλ ψ0 i| + |hf, Lj−1 Tλ ψ` i|2 . (5.2)
λ∈Λ `=1 λ∈Λ

j−1 T ψ i|2 ,
P
Applying Lemma 4.2 on λ∈Λ |hf, L λ 0 we get

X X n X
X
j−1 2 j−2 2
|hf, L Tλ ψ0 i| = |hf, L Tλ ψ0 i| + |hf, Lj−2 Tλ ψ` i|2 . (5.3)
λ∈Λ λ∈Λ `=1 λ∈Λ

By (5.2) and (5.3), we have

X X X j−1
n X X
j 2 j−2 2
|hf, L Tλ ψ0 i| = |hf, L Tλ ψ0 i| + |hf, Lp Tλ ψ` i|2 .
λ∈Λ λ∈Λ `=1 λ∈Λ p=j−2

Repeating the above arguments, for any m < j, we have

X X X j−1
n X X
|hf, Lj Tλ ψ0 i|2 = |hf, Lm Tλ ψ0 i|2 + |hf, Lp Tλ ψ` i|2 . (5.4)
λ∈Λ λ∈Λ `=1 λ∈Λ p=m

It follows from (5.1) and (5.4) that for all j ≥ J and for all m < j,

X X j−1
n X X
2 m 2
(1 − )kf k ≤ |hf, L Tλ ψ0 i| + |hf, Lp Tλ ψ` i|2 ≤ (1 + )kf k2 .
λ∈Λ `=1 λ∈Λ p=m

Letting m → −∞ in above and using (ii) of Lemma 4.3, we have

j−1
n X X
X
2
(1 − )kf k ≤ |hf, Lp Tλ ψ` i|2 ≤ (1 + )kf k2 . (5.5)
`=1 λ∈Λ p=−∞
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 89

Letting j → ∞ in (5.5), we have

∞
n X X
X
2
(1 − )kf k ≤ |hf, Lp Tλ ψ` i|2 ≤ (1 + )kf k2 .
`=1 λ∈Λ p=−∞

Since  > 0 is arbitrary, we obtain

n XX
X
|hf, Lp Tλ ψ` i|2 = kf k2 (5.6)
`=1 λ∈Λ p∈Z

Now, since (5.6) holds on the dense subset of L2 (R), then, by Lemma 2.2, it will
hold on L2 (R), which completes the proof.

The next theorem gives the generalized (or oblique) extension principle for
nonuniform wavelet frames in L2 (R). It gives a more flexible technique to con-
struct nonuniform wavelet frames.

Theorem 5.2. Let {ψ` , H` }n`=0 be a nonuniform general setup. Assume that
there exists a strictly positive function θ ∈ L∞ (R) for which

lim θ(γ) = 1,
γ→0+

and
n
X
2
θ(2N γ)|H0 (γ)| + |H` (γ)|2 = θ(γ).
`=1

Then {Lj Tλ ψ` } j∈Z,λ∈Λ is a Parseval nonuniform wavelet frame for L2 (R).

`=1,2,··· ,n

Proof. Define ψe0 ∈ L2 (R) such that

p
e (γ) =
ψ θ(γ)ψb0 (γ). (5.7)
b
0

Define the functions He0, H

e1, . . . , H
e n as follows:
s s
θ(2N γ) 1
H
e 0 (γ) = H0 (γ), H e ` (γ) = H` (γ), ` = 1, 2, . . . , n.
θ(γ) θ(γ)

Then we have
p p
ψ
e (2N γ) = θ(2N γ)ψb0 (2N γ) = θ(2N γ)H0 (γ)ψc0 (γ)
b
0
  s
ψe (γ)  θ(2N γ)
b
= θ(2N γ) H0 (γ) p0
p
= H0 (γ)ψ
e (γ)
b
0
θ(γ) θ(γ)

=H
e 0 (γ)ψ
e (γ) (5.8)
b
0
90 Hari Krishan Malhotra and Lalit Kumar Vashisht

and
p
lim ψ
e (γ) = lim θ(γ)ψb0 (γ) = 1. (5.9)
b
0
γ→0+ γ→0+

Since {ψ` , H` }n`=0 is a nonuniform general setup, by (5.7), we have

 
1
Supp ψ 0 (γ) ⊆ 0, (5.10)
b
e
4N
and
n
X n
X
e ` (γ)|2 = |H
|H e 0 (γ)|2 + e ` (γ)|2
|H
`=0 `=1
n
X |H` (γ)|2
θ(2N γ) 1
= |H0 (γ)|2 + = θ(γ) = 1. (5.11)
θ(γ) θ(γ) θ(γ)
`=1

Thus,
e ` (γ) ∈ L∞ (R) for ` = 0, 1, . . . , n.
H (5.12)

Let ψe1 , ψe2 , . . . , ψen ∈ L2 (R) be such that

ψ
e (2N γ) = H
e ` (γ)ψ
e (γ), ` = 1, . . . , n. (5.13)
b b
` 0

Define
e 
H0 (γ)
H
 1 (γ) 
e 
H(γ)
e = .  .
 .. 
H
e n (γ)
(n+1)×1

e ` }n is a nonuniform
Then, by (5.8), (5.9), (5.10) and (5.12), the collection {ψe` , H `=0
general setup.
Using (5.11), we have
" n #
X
e ∗ H(γ)
H(γ) e = |He ` (γ)|2 = 1.
`=0

Hence, by Theorem 5.1, {Lj Tλ ψe` } j∈Z,λ∈Λ is a Parseval nonuniform wavelet frame
`=1,2,...,n
for L2 (R).
Next, we compute
 
  ψe (γ)
b
e ` (γ) θ(γ)  p0
p
ψb` (2N γ) = H` (γ)ψb0 (γ) = H =H
e ` (γ)ψ
e (γ) = ψ
e (2N γ).
b b
0 `
θ(γ)

This gives ψ` = ψe` . Hence, the system {Lj Tλ ψ` } j∈Z,λ∈Λ is a Parseval nonuniform
`=1,2,··· ,n
wavelet frame for L2 (R).
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 91

Remark 5.3. It is worth noticing that, when θ = 1, Theorem 5.1 can be

obtained from Theorem 5.2.
Remark 5.4. All the results will also remain true if we replace r/N by an
arbitrary a ∈ (0, 2). This is pointed out by the referee.

6. Examples
This section gives some applicative examples of the UEP and its generalized
version. The example below illustrates Theorem 5.1.
Example 6.1. Let N = 2, r = 3, 0 6= a ∈ R, and ψ0 ∈ L2 (R) be such that
sin(aγ)
ψb0 (γ) = χ]0,1/8] (γ).
aγ
Then
(i) lim ψb0 (γ) = 1;
γ→0+
 
1
(ii) Supp ψ0 ⊆ 0,
c ; and
8
sin(4aγ)
(iii) ψb0 (4γ) = χ]0,1/8] (4γ)
4aγ
4 sin(aγ) cos(aγ) cos(2aγ)
= χ]0,1/32] (γ)χ]0,1/8] (γ) = H0 (γ)ψb0 (γ),
4aγ
where H0 (γ) = cos(aγ) cos(2aγ)χ]0,1/32] (γ).
Let
H1 (γ) = cos(2aγ) sin(aγ)χ]0,1/32] (γ),
H2 (γ) = sin(2aγ)χ]0,1/32] (γ),
H3 (γ) = χR\]0,1/32] (γ).
Let ψ1 , ψ2 , ψ3 ∈ L2 (R) be such that
ψb` (4γ) = H` (γ)ψb0 (γ), ` = 1, 2, 3.
Choose
 
H0 (γ)
H1 (γ)
H(γ) = 
H2 (γ) .


H3 (γ)

Then {ψ` , H` }3`=0 is a nonuniform general setup such that

H(γ)∗ H(γ) = |H0 (γ)|2 + |H1 (γ)|2 + |H2 (γ)|2 + |H3 (γ)|2 = 1.
 

Hence, by Theorem 5.1, {Lj Tλ ψ` }j∈Z,λ∈{0,3/2}+2Z is a nonuniform Parseval

`=1,2,3
wavelet frame for L2 (R).
92 Hari Krishan Malhotra and Lalit Kumar Vashisht

To conclude the paper, we illustrate Theorem 5.2 with the following example.
Example 6.2. Let N = 2, r = 3 and ψ0 ∈ L2 (R) be such that for any fixed
t ∈ R,

ψb0 (γ) = eitγ χ[0,1/8] (γ).

Then
(i) lim ψb0 (γ) = 1;
γ→0+
 
1
(ii) Supp ψ0 (γ) ⊆ 0,
b ; and
8

(iii) ψ0 (4γ) = e4itγ χ[0,1/8] (4γ) = e4itγ χ[0,1/32] (γ)χ[0,1/8] (γ) = H0 (γ)ψb0 (γ),
where H0 (γ) = e3itγ χ[0,1/32] (γ) ∈ L∞ (R).
Let θ(γ) = 1, and define H1 (γ) = χR\[0,1/32] . Then the collection {ψ` , H` }1`=0
is a nonuniform general setup such that

θ(4γ)|H0 (γ)|2 + |H1 (γ)|2 = θ(γ).

Hence, by Theorem 5.2, the nonuniform wavelet system {Lj Tλ ψ1 }j∈Z,λ∈{0,3/2}+2Z

is a Parseval frame for L2 (R).

Acknowledgments. The authors are deeply indebted to the anonymous

referee for careful reading the manuscript and for offering valuable comments
and detailed suggestions which enabled the authors to substantially improve the
paper. We are also thankful to the referee for providing short proof of Lemma
4.1 and Remark 5.4, which says that all the results can be generalized by taking
an arbitrary a ∈ (0, 2) in place of r/N .
The first author is supported by the University Grants Commission (UGC),
India. Grant No.: 19/06/2016(i)EU-V.

References
[1] J. Benedetto and O. Treiber, Wavelet Frames: Multiresolution Analysis and Exten-
sion Principles, Birkhäuser, Boston, 2001, 1–36.
[2] A. Boggess and F.J. Narcowich, A First Course in Wavelets with Fourier Analysis,
John Wiley & Sons, Inc., Hoboken, NJ, 2009.
[3] P.G. Casazza and G. Kutyniok, Finite Frames: Theory and Applications,
Birkhäuser, 2012.
[4] O. Christensen, H.O. Kim, and R.Y. Kim, Extensions of Bessel sequences to dual
pairs of frames, Appl. Comput. Harmon. Anal. 34 (2013), No 2, 224–233.
[5] O. Christensen, An Introduction to Frames and Riesz Bases, 2nd ed., Birkhäuser,
2016.
[6] O. Christensen and S.S. Goh, The unitary extension principle on locally compact
abelian groups, Appl. Comput. Harmon.Anal. 47 (2019), No. 1, 1–29.
 Unitary Extension Principle for Nonuniform Wavelet Frames in L2 (R) 93

[7] C. K. Chui, An Introduction to Wavelets, Academic Press, Inc., Boston, 1992.

[8] I. Daubechies, Ten Lectures on Wavelets, SIAM, Philadelphia, 1992.
[9] I. Daubechies, B. Han, A. Ron, and Z. Shen, Framelets: MRA-based constructions
of wavelet frames, Appl. Comput. Harmon. Anal. 14 (2003), No 1, 1–46.
[10] Deepshikha and L.K. Vashisht, A note on discrete frames of translates in CN ,
TWMS J. Appl. Eng. Math. 6 (2016), No. 1, 143–149.
[11] Deepshikha and L.K. Vashisht, Necessary and sufficient conditions for discrete
wavelet frames in CN , J. Geom. Phys. 117 (2017), 134–143.
[12] Dao-Xin Ding, Generalized continuous frames constructed by using an iterated func-
tion system, J. Geom. Phys. 61 (2011) 1045–1050.
[13] R.J. Duffin and A.C. Schaeffer, A class of nonharmonic Fourier series, Trans. Amer.
Math. Soc.72 (1952), 341–366.
[14] J.P. Gabardo and M.Z. Nashed, Nonuniform multiresolution analysis and spectral
pairs, J. Funct. Anal. 158 (1998), 209–241.
[15] J.P. Gabardo and X. Yu, Wavelets associated with nonuniform multiresolution anal-
yses and one-dimensional spectral pairs, J. Math. Anal. Appl. 323 (2006), 798–817.
[16] B. Han, Framelets and Wavelets: Algorithms, Analysis, and Applications,
Birkhäuser, 2017.
[17] C. Heil, A Basis Theory Primer, Expanded edition, Birkhäuser, 2011.
[18] E. Hernandez and G. Weiss, A First Course on Wavelets, CRC Press, Boca Raton,
1996.
[19] A. Krivoshein, V. Protasov, and M. Skopina, Multivariate Wavelet Frames, Springer,
2016.
[20] A. Ron and Z. Shen, Affine systems in L2 (Rd ): The analysis of the analysis operator,
J. Funct. Anal. 148 (1997), 408–447.
[21] M.B. Ruskai, G. Beylkin, R. Coifman, I. Daubechies, S. Mallat, Y. Meyer, and
L. Raphael, Wavelets and Their Applications, Jones and Bartlett Publishers,
Boston, MA, 1992.
[22] L.K. Vashisht and Deepshikha, Weaving properties of generalized continuous frames
generated by an iterated function system, J. Geom. Phys. 110 (2016), 282–295.
[23] X. Yu and J.P. Gabardo, Nonuniform wavelets and wavelet sets related to one-
dimensional spectral pairs, J. Approx. Theory 145 (2007), No. 1, 133–139.
[24] R.A. Zalik, Riesz bases and multiresolution analyses, Appl. Comput. Harmon. Anal.
7 (1999), No. 3, 315–331.
[25] R.A. Zalik, Orthonormal wavelet systems and multiresolution analyses, J. Appl.
Funct. Anal. 5 (2010), No. 1, 31–41.

Received January 9,2020, revised March 29, 2020.

Hari Krishan Malhotra,

Department of Mathematics, University of Delhi, Delhi-110007, India,
E-mail: [email protected]
 94 Hari Krishan Malhotra and Lalit Kumar Vashisht

Lalit Kumar Vashisht,

Department of Mathematics, University of Delhi, Delhi-110007, India,
E-mail: [email protected]

Ïðèíöèï óíiòàðíîãî ðîçøèðåííÿ äëÿ íåîäíîðiäíèõ

âåéâëåò-ôðåéìiâ â L2 (R)
Hari Krishan Malhotra and Lalit Kumar Vashisht
Ôðåéìè Ïàðñåâàëÿ ïðèâåðíóëè óâàãó iíæåíåðiâ i ôiçèêiâ çàâäÿêè ¨õ
ïîòåíöiéíîìó çàñòîñóâàííþ â îáðîáöi ñèãíàëiâ. Ó öié ðîáîòi ìè âèâ÷à¹ìî
ïîáóäîâó íåîäíîðiäíèõ âåéâëåò-ôðåéìiâ Ïàðñåâàëÿ äëÿ ïðîñòîðó Ëåáå-
ãà L2 (R), äå âiäïîâiäíà ìíîæèíà çñóâiâ íå îáîâ'ÿçêîâî ì๠áóòè ãðóïîþ.
Îñíîâíà ìåòà äàíî¨ ðîáîòè  äîâåñòè ïðèíöèï óíiòàðíîãî ðîçøèðåííÿ
(ÏÓP) òà ïðèíöèï êîñîãî ðîçøèðåííÿ (ÏÊÐ) äëÿ ïîáóäîâè ìóëüòèãå-
íåðîâàíèõ íåîäíîðiäíèõ âåéâëåò-ôðåéìiâ Ïàðñåâàëÿ äëÿ L2 (R). Òàêîæ
íàâåäåíî äåÿêi ïðèêëàäè, ùî iëþñòðóþòü ðåçóëüòàòè.
Êëþ÷îâi ñëîâà: ôðåéì Ãiëüáåðòà, íåîäíîðiäíà âåéâëåò-ñèñòåìà,
ïðèíöèï óíiòàðíîãî ðîçøèðåííÿ

View publication stats