Continuous abstract wavelet transform on homogeneous space
Jyoti Sharma, Ajay Kumar

TL;DR
This paper investigates the properties of the continuous wavelet transform on homogeneous spaces, revealing its infinite measure support, approximation capabilities, and an analogue of the Heisenberg inequality.
Contribution
It introduces new theoretical results on wavelet transforms on homogeneous spaces, including measure properties, approximation, and inequalities, expanding understanding of wavelet analysis in this context.
Findings
Wavelet transform support has infinite measure.
Pointwise homogeneous approximation property established.
An analogue of Heisenberg inequality derived for wavelet transform.
Abstract
The support of wavelet transform associated with square integrable irreducible representation of a homogeneous space is shown to have infinite measure. Pointwise homogeneous approximation property for wavelet transform has been investigated. An analogue of Heisenberg type inequality has been also obtained for wavelet transform
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsMathematical Analysis and Transform Methods · Image and Signal Denoising Methods · Seismic Imaging and Inversion Techniques
Continuous Abstract Wavelet transform on Homogeneous Spaces
JYOTI SHARMA
Department of Mathematics, University of Delhi, Delhi, 110007, India.
and
AJAY KUMAR∗
Department of Mathematics, University of Delhi, Delhi, 110007, India.
Abstract.
The support of wavelet transform associated with square integrable irreducible representation of a homogeneous space is shown to have infinite measure. Pointwise homogeneous approximation property for wavelet transform has been investigated. An analogue of Heisenberg type inequality has been also obtained for wavelet transform.
Key words and phrases:
Abstract wavelet transform, square integrable representation, Heisenberg type inequality
2010 Mathematics Subject Classification:
Primary 43A85; Secondary 65T60; 42C40
∗Corresponding author
1. Introduction
Fourier transform is used to analyse the frequency property of a given signal. However, due to loss of information about time, other transforms like Gabor transform and wavelet transform have been found to be more useful. Continuous wavelet transform has been widely used in signal and image processing for investigating time-varying frequency. Unlike Fourier analysis, wavelet analysis expands functions not in terms of trigonometric polynomials but in terms of wavelets, which are generated in the form of translations and dilations of a fixed function called the admissible wavelet. Wavelets obtained in this way have special scaling properties. They are localized in time and frequency, permitting a closer connection between the function being represented and their coefficients. These have been recently studied in harmonic analyis (see [1, 3, 8, 5, 13]).
We begin by defining wavelet transform for an arbitrary homogeneous space. Let G be a locally compact group with left Haar measure and H a closed subgroup of . The left coset space is a homogeneous space with quotient topology. Also, acts on via . We assume that for all where and are the modular function of and respectively. For the pair a rho function is a continuous function such that for all and (see [7, p. 65]). Let be the group consisting of all unitary operators on some Hilbert space . A continuous unitary representation of the homogeneous space is a map from into for which is a continuous map from such that for each and ,
[TABLE]
Let be a square integrable representation of , i.e. there exists some and , satisfying
[TABLE]
Then, satisfying (1.1) is called an admissible wavelet. A continuous wavelet transform associated to is a linear operator defined by
[TABLE]
If is irreducible, then is a bounded linear operator from Hilbert space and where is a relatively -invariant measure on which arises from the rho function Also, is a reproducing kernel Hilbert space with pointwise bounded kernel and the operator is an isometry. For detailed study of wavelet transform on homogeneous space, one can refer to [5].
In this paper, we first show that for the support of is a set of infinite measure. We prove that every pair of admissible vectors possesses homogeneous approximation property. In addition, we study the wavelet groups of the form where are locally compact, type I groups. Moreover, lower estimate on norm of wavelet transform and Heisenberg type inequality have been obtained.
2. Concentration of Wavelet transform
Throughout this section, we assume that is compact and for all . Thus, is a -invariant measure, i.e. for all and . To investigate the support of the wavelet transform , we begin with following lemmas.
Lemma 2.1**.**
If are such that then the function such that , is continuous.
Proof.
Let be the space of continuous functions with compact support. Since and is dense in therefore we choose such that , where is arbitrary. Then, we consider
[TABLE]
The map is continuous from [12, Proposition 1.25], where for all and . Hence, is a continuous function. ∎
In the next lemma, we generalize [10, Lemma 2.1] to homogeneous spaces.
Lemma 2.2**.**
Let , the identity component of be non-compact and be such that Then for there exists satisfying
[TABLE]
Proof.
Define by Since is -invariant, therefore Then by Lemma 2.1, is a continuous function. Also is regular, so there exists a compact subset of such that Let , where is a compact subset of . Thus, is a compact subset of . Choose For such a choice of we have . Moreover,
[TABLE]
Therefore, it follows that
[TABLE]
Again, we have
[TABLE]
Thus, is a non-constant continuous function on the connected set . So, we can choose such that
[TABLE]
In the next result, we prove that the support of wavelet transform has infinite measure.
Theorem 2.3**.**
Let and be as in Lemma 2.2 and be an admissible wavelet associated with an irreducible square integrable representation . Then for any , the set has infinite measure.
Proof.
It is enough to show that for every subset of with , the set . For this assume that and with Let . Then,
[TABLE]
By Lemma 2.2, there exists satisfying
[TABLE]
Take and and again using Lemma 2.2, there exists satisfying
[TABLE]
Continuing in the same manner, we get an increasing sequence of subsets of given by
[TABLE]
where for all and satisfy
[TABLE]
Define Then,
[TABLE]
Consider the family of functions on as follows
[TABLE]
We will show by induction that for all Clearly, Let us assume that for some Then,
[TABLE]
which implies that Using (2.1), we conclude that the family of functions on is linearly independent vanishing outside the set of finite measure. Since is a reproducing kernel Hilbert space with point-wise bounded kernel, therefore each subspace of consisting of functions having support on a set of finite measure must be of finite dimension [16]. If such that then a.e. and hence Thus, the support of is a set of infinite measure for every non-zero . ∎
Remarks 2.4**.**
- (1)
If we drop the irreducibility of , then the conclusion of Theorem 2.3 may not hold. Consider a locally compact abelian topological group and the left regular representation of . Then every non-zero function is an admissible wavelet. Fix non-zero functions Then, , where is the convolution and . Hence 2. (2)
If G is an abelian locally compact group and is an irreducible representation of , then the support of wavelet transform is In this case , where is the character of associated with This may not be true if we consider a non-abelian compact group G. Consider Let be an irreducible unitary representation of G given by:
[TABLE]
where , is fixed and being the outer tensor product. If we consider , then Thus, which implies that
[TABLE]
where is the cardinality of
We conclude this section section by giving examples of group having a square integrable irreducible representation.
Examples 2.5**.**
-
Groups of the type where is a closed subgroup of
-
(i)
Let , where .
A square integrable irreducible representation of on is given by
[TABLE]
Also, is admissible if and only if C_{\psi}=\int_{\mathbb{R}^{2}}\Big{|}\frac{\widehat{\psi}(\gamma_{1},\gamma_{2})}{\sqrt{|\gamma_{1}\gamma_{2}|}}\Big{|}^{2}d\gamma_{1}d\gamma_{2} [3]. 2. (ii)
where .
A square integrable irreducible representation on is given by
[TABLE]
Also, is admissible if and only if 3. (iii)
Shear group with group operation where
[TABLE]
A square integrable irreducible representation on is given by
[TABLE]
and function is admissible if and only if [4].
-
Low dimensional nilpotent Lie groups.
-
(i)
Reduced Weyl Heisenberg group, with the group operation An irreducible square integrable representation on the Hilbert space is given by
[TABLE]
Every is an admissible wavelet and .
For the group structure of 5-dimensional groups given below (see [14]).
- (ii)
an irreducible square integrable representation on the Hilbert space is given by
[TABLE]
where Every is an admissible wavelet and . 2. (iii)
a square integrable irreducible representation on the Hilbert space is given by
[TABLE]
where Every is an admissible wavelet and . 3. (iv)
an irreducible square integrable representation on the Hilbert space is given by
[TABLE]
where Every is an admissible wavelet and .
- Let be a locally compact group with non-compact identity component having a square integrable irreducible representation and be a compact group of the form . Let be an irreducible representation of which is identity on . Then, has an irreducible square integrable representation , where being the outer tensor product [7, Theorem 7.17].
3. Homogeneous Approximation Property
In this section, we deal with homogeneous approximation property for the wavelet transform on homogeneous space where is a closed subgroup of . For an irreducible unitary representation of with representation space , a pair consisting of admissible wavelets in is said to be an admissible pair if and
[TABLE]
is a non-zero scalar. As proved in [5, Theorem 2.1], for admissible wavelets , the operator intertwines with . Thus, there exists a scalar such that , where is the identity operator. Therefore, we have
[TABLE]
for all . Taking , we have
[TABLE]
If is an admissible pair, then which further implies that every can be decomposed as
[TABLE]
where the integral is defined in the weak sense.
Definition 3.1**.**
An admissible pair is said to have homogeneous approximation property in , if for any and there exists some compact subset of such that every compact subset of with satisfies
[TABLE]
for all
Theorem 3.2**.**
Let be a -compact homogeneous space with an irreducible square integrable representation . Then every admissible pair possesses homogeneous approximation property in .
Proof.
Let be an arbitrary compact subset of and be a compact subset of such that . Then for and we have
[TABLE]
But, is -compact, therefore, there exists a sequence of compact subsets of such that for all and . Since , as and , therefore by Lebesgue dominated convergence theorem, it follows that
[TABLE]
Thus, there exists a such that
[TABLE]
or, equivalently,
[TABLE]
Therefore, we can choose a compact subset of such that
[TABLE]
for every compact subset of containing and every ∎
We now consider similitude group with group operation
[TABLE]
The left Haar measure on is given by The mapping defined by
[TABLE]
is a square integrable irreducible representation of . For ,
[TABLE]
is independent of almost every (see [1]). Also, for an admissible pair ,
[TABLE]
is independent of upto a measure zero set.
In the next result, we further deal with the point-wise homogeneous approximation property for wavelet transform on similitude group. The idea of results is motivated by [13] and [15].
Theorem 3.3**.**
Let be an admissible pair in . For and define
[TABLE]
Then, and
[TABLE]
Proof.
First we show that is well defined. For this, we compute
[TABLE]
Let be arbitrary. Then
[TABLE]
Thus, using Tonneli theorem, the function is integrable. Again, we have
[TABLE]
Thus,
[TABLE]
Therefore, by Riesz representation theorem Now on using Plancherel Theorem, we have
[TABLE]
Since , therefore we have
[TABLE]
Theorem 3.4**.**
Let be an admissible pair in and with . Then for any , there exist such that for any with and , the following holds
[TABLE]
Proof.
For arbitrary and with , define
[TABLE]
and
[TABLE]
Then,
[TABLE]
Since both and are admissible, therefore we get
[TABLE]
and
[TABLE]
Also,
[TABLE]
and
[TABLE]
Thus,
[TABLE]
By Lebesgue dominated convergence theorem, we get
[TABLE]
and
[TABLE]
Therefore, we can choose such that for any with and , the conclusion of the theorem holds. ∎
4. Wavelet group of the form
Let and be locally compact, second countable, unimodular, group of type I and be a map from to where is a group of automorphism of . Let with group operation . The left Haar measure on is given by , where is a positive homomorphism of satisfying Define a representation of on Hilbert space such that
[TABLE]
A function is said to be a feasible wavelet, if there exists a constant such that for almost every and for all
[TABLE]
It may be noted that if is a locally compact abelian group, then for a feasible wavelet ,
[TABLE]
is independent of almost every For the case when is a locally compact abelian group, one may refer to [6]. The continuous wavelet transform with respect to a feasible wavelet is an operator defined as Also,
[TABLE]
where . Using feasibility condition and Plancherel’s theorem [7, Theorem 7.36] for locally compact, unimodular, type I groups, one can show the following result:
Lemma 4.1**.**
Let be a feasible wavelet and . Then .
In fact, one can show that is a reproducing kernel Hilbert space with point-wise bounded kernel given by
[TABLE]
Proceeding as in Lemma 2.2 and Theorem 2.3, we obtain the following:
Theorem 4.2**.**
Let be an abstract wavelet group with non compact and be a feasible wavelet. If , then the set has infinite measure.
Theorem 4.3**.**
Let be an abstract wavelet group and be a feasible wavelet. Let be such that Then for any ,
[TABLE]
Proof.
Let be the orthogonal projection from to the closed subspace and be the orthogonal projection from to the closed subspace of functions with support contained in Now,
[TABLE]
Now being a projection on a reproducing kernel Hilbert space can be represented by
[TABLE]
where and is the reproducing kernel. Thus, for every we have
[TABLE]
From [9], the operator norm is given by
[TABLE]
On substituting the value into (4), we get the desired inequality. ∎
Next, we prove an analogue of Heisenberg type inequality for wavelet transform.
Theorem 4.4**.**
Let be a wavelet group, where is a compact group and be a feasible wavelet. Then for any , we have
[TABLE]
Proof.
For almost all , using Heisenberg inequality for Fourier transform on (see [2]), we have
[TABLE]
On integrating both sides with respect to the measure , we get
[TABLE]
Using Lemma 4.1 and Cauchy-Schwarz inequality, it follows that
[TABLE]
Since,
[TABLE]
therefore, on using feasibility condition, we obtain
[TABLE]
On substituting (4) into (4), we get
[TABLE]
Remark 4.5**.**
- (i)
Using the data given in [2], one may write the explicit form of the Heisenberg type inequality when is a locally compact abelian group with non-compact identity component or -dimensional nilpotent Lie group or of the form where is compact subgroup of group of automorphism of . 2. (ii)
As a particular case of above results, one can obtain results for abstract shearlet transform (see [11]) by taking , where and K are locally compact groups and Aut is a homomorphism.
Acknowledgements
The first author is supported by UGC under joint UGC-CSIR Junior Research Fellowship (Ref. No:21/12/2014(ii)EU-V).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] S. T. Ali, J.-P. Antoine and J.-P. Gazeau, Coherent states, Wavelets, and their Generalizations , 2nd edn., Springer, 2014.
- 2[2] A. Bansal and A. Kumar, Generalized analogs of the Heisenberg uncertainty inequality , J. Inequal. Appl. 168 (2015), 15pp.
- 3[3] D. Bernier and K. F. Taylor, Wavelets from square-integrable representations , Siam J. Math. Anal. 27 (1996), no. 2, 594-608.
- 4[4] S. Dahlke, G. Steidl and G. Teschke, The continuous shearlet transform in arbitrary space dimensions , J. Fourier Anal. Appl. 16 (2010), no. 3 , 340-364.
- 5[5] F. Esmaeelzadeh, R. A. Kamyabi-Gol and R. Raisi Tousi, On the continuous wavelet transform on homogeneous space , Int. J. of Wavelets Multiresolut. and Inf. Process. 10 (2012), no. 4, 1250038, 18 pp.
- 6[6] M. Fashandi and R. A. Kamyabi Gol, Continuous wavelet transform on a special homogeneous space , J. Math. Phys. 44 (2003), 4260–4266.
- 7[7] G. B. Folland, A course in abstract harmonic analysis , CRC Press, 2016.
- 8[8] H. F u ¨ ¨ 𝑢 \ddot{u} hr, Abstract Harmonic Analysis of Continuous Wavelet Transforms , Lecture Notes in Mathematics, 1863 Springer, Berlin, 2005.
