Next: The Pyramid Algorithm
Up: Multiresolution Analysis
Previous: Multiscale Analysis as a
Contents
Index
Multiscale Analysis vs Multiresolution Analysis: MSA or MRA?
The two names ``multiscale analysis'' (=MSA) and ``multiresolution
analysis''(=MRA) refer to the same concept. Both are characterized by
the discrete set of powers of the number 2,
and the corresponding set of orthonormal basis functions
The difference is that MRA and MSA highlight different aspects of the
same thing. As
increases, the scale increases but the resolution
decreases. This is like stepping away from a picture.
Consider the array of functions
This array is a set of identical localized graphs, each one displaced
by the amount
from its nearest neighbor. Thus each of these
graphs serves as marker on the real line and
is the distance
between successive makers. In brief, the real line equipped with this
set of markers constitutes a new kind of measuring rod for measuring
signals. The integer
specifies the nature - the resolution - of
this measuring rod. Every integral increase in
increases
(decreases) the scale for performing these mesurements, and hence
decreases (increases) the resolution of the measuring rod.
Figure 2.24 depicts several such measuring rods.
Figure 2.24:
Nine different measuring rods. Each is graduated with its own
set of markers, i.e. shifted scaling functions
,
ranging from a set of very high resolution (
) markers,
through a set of medium resolution (
), to the set of lowest
resolution (
) markers. A high resolution measuring rod
accomodates additional high resolution markers, which are, however,
not depicted in this figure. The markers of each rod are uniformly
spaced, as they must. The novelty of these rods is that each marker
has the mathematically precise internal structure of a
wavepacket. This novelty permits one to measure not only the locations
of specific features in a given signal but also their amplitudes.
 |
Next: The Pyramid Algorithm
Up: Multiresolution Analysis
Previous: Multiscale Analysis as a
Contents
Index
Ulrich Gerlach
2007-04-05