Discrete and Continuous Wavelet Transformationson the Heisenberg GroupAzita MayeliTechnische Universit¨at Munc¨ hen2005¨ ¨Zentrum Mathematik der Technischen Universitat MunchenLehrstuhl Univ.-Prof. Dr. J. ScheurleDiscrete and Continuous Wavelet Transformationson the Heisenberg GroupAzita MayeliVollst¨andiger Abdruck der von der Fakult¨at fur¨ Mathematik der Technis-chen Universit¨at Munc¨ hen zur Erlangung des akademischen Grades einesDoktors der Naturwissenschaften (Dr. rer. nat.)genehmigten Dissertation.Vorsitzender: Univ.-Prof. Dr. Jur¨ gen ScheurlePrufer¨ der Dissertation: 1. apl.-Prof. Dr. Gun¨ ter Schlichting2. Priv.-Doz.Dr. Hartmut Fuhr¨3. Prof. Daryl N.GellerState University of New York/ USA(schriftliche Beurteilung)Die Dissertation wurde am 10. November 2005 bei der Technischen Universit¨at Munc¨ heneingereicht und durch die Fakultat¨ fur¨ Mathematik am 22. April 2006 angenommen.AcknowledgementsI would like to gratefully acknowledge the enthusiastic supervision of Professor Gun¨ terSchlichting during this work, who made this work possible, feasible and pleasurable inrough order of appearance in my life. I also want to thank him for giving me the oppor-tunity to travel and getting in touch with distinguished scientists in the field of HarmonicAnalysis.Formost, I wish to express my deep gratitude to PD. Dr.
Wavelet analysis is a still developing area in the mathematical sciences.
Already early
in the development of the wavelets both the discrete and the continuous transformation
were examined.
The main aim of the theory of wavelet analysis is to find convenient ways to decompose a
given function into elementary building blocks. Historically, the Haar basis, constructed
in 1910 long before the term “wavelet” was created, was the first orthonormal wavelet basis inL2(R). But it was only recently discovered that the construction works because of
ered the first continuous orthogonal wavelets. His wavelets have exponential decay and were inCk,kite.Thenarybutfinurtcoi,nxectnotsrartibtrfStoenndpedeinsaw,grebmo¨
the Meyer wavelet [40]. The images of the Meyer wavelets under the Fourier transform were compactly supported and were inCk(kmay be∞). With the notion of multires-
olution analysis, introduced by Mallat [38] and Meyer [41], a systematic framework for
understanding these orthogonal expansions was developed, see for example [38] and [41]
for details. This framework gave a satisfactory explanation for all these constructions,
and provided a tool for the construction of other bases. Thus, multiresolution analysis is an important mathematical tool to understand and construct a wavelet basis ofL2(R),
i.e., a basis that consists of the scaled and integer translated versions of a finite number
1
2
of functions.
Preface
In recent years, multiresolution analysis for the Euclidean groupRhas received extensive
investigation. Also, various extensions and generalizations were considered. There are
literally hundreds of sources dealing with this connection. In [37] multiresolution analy-sis forRn [8] Dahlkewhose scaling functions are characteristic functions are considered. extended multiresolution analysis to abelian locally compact groups. Baggett, et al. [2]
considered the existence of wavelets in general Hilbert space based on the formulation of
multiresolution analysis by using an abstract approach.
An alternative construction, imposing less restrictions on the wavelet functions, is the
continuous wavelet transform. Thecontinuous wavelet transformation can be interpreted
as a phase space representation. Their filters and approximation characteristics have been
examined. The group-theoretical approach allows a simple generalization for instance of
wavelet transformation to high-dimensions Euclidean space (see [18]) or more general sit-uations. Wavelet transformation in several dimension , exactly as in one dimension, may be derived from the similitude group ofRn(n >1), consisting of dilations, rotations and
translations. Of course, the most interesting case of applications isn= 2, where wavelets
have become a useful tool in image processing.
The construction of generalized continuous wavelet transform is investigated in the frame-
work of irreducible, square-integrable representations of locally compact groups. The
square integrability of representations guarantees the existence of a so-called admissible
vector and an inverse wavelet transform [3], [28]. General existence theorems for square-
integrable representation can be found in [12]. The existence of admissible vectors can
also be considered when the irreducibility requirement can be dropped, as for example in
[19], using a connection between generalized wavelet transforms and Plancherel theory.
Introduction to the Wavelets onR
The wavelet transform of a function onR, a signal so-called, depends on two variables: