C09CCF (PDF version)
C09 Chapter Contents
C09 Chapter Introduction
NAG Library Manual

NAG Library Routine Document


Note:  before using this routine, please read the Users' Note for your implementation to check the interpretation of bold italicised terms and other implementation-dependent details.

+ Contents

    1  Purpose
    7  Accuracy

1  Purpose

C09CCF computes the one-dimensional multi-level discrete wavelet transform (DWT). The initialization routine C09AAF must be called first to set up the DWT options.

2  Specification

REAL (KIND=nag_wp)  X(N), C(LENC)

3  Description

C09CCF computes the multi-level DWT of one-dimensional data. For a given wavelet and end extension method, C09CCF will compute a multi-level transform of a data array, xi, for i=1,2,,n, using a specified number, nl, of levels. The number of levels specified, nl, must be no more than the value lmax returned in NWL by the initialization routine C09AAF for the given problem. The transform is returned as a set of coefficients for the different levels (packed into a single array) and a representation of the multi-level structure.
The notation used here assigns level 0 to the input dataset, x, with level 1 being the first set of coefficients computed, with the detail coefficients, d1, being stored while the approximation coefficients, a1, are used as the input to a repeat of the wavelet transform. This process is continued until, at level nl, both the detail coefficients, dnl, and the approximation coefficients, anl are retained. The output array, C, stores these sets of coefficients in reverse order, starting with anl followed by dnl,dnl-1,,d1.

4  References


5  Parameters

1:     N – INTEGERInput
On entry: the number of elements, n, in the data array x.
Constraint: this must be the same as the value N passed to the initialization routine C09AAF.
2:     X(N) – REAL (KIND=nag_wp) arrayInput
On entry: X contains the one-dimensional input dataset xi, for i=1,2,,n.
3:     LENC – INTEGERInput
On entry: the dimension of the array C as declared in the (sub)program from which C09CCF is called. C must be large enough to contain the number, nc, of wavelet coefficients. The maximum value of nc is returned in NWC by the call to the initialization routine C09AAF and corresponds to the DWT being continued for the maximum number of levels possible for the given data set. When the number of levels, nl, is chosen to be less than the maximum, then nc is correspondingly smaller and LENC can be reduced by noting that the number of coefficients at each level is given by n-/2 for MODE='P' in C09AAF and n-+nf-1/2 for MODE='H', 'W' or 'Z', where n- is the number of input data at that level and nf is the filter length provided by the call to C09AAF. At the final level the storage is doubled to contain the set of approximation coefficients.
Constraint: LENCnc, where nc is the number of approximation and detail coefficients that correspond to a transform with NWL levels.
4:     C(LENC) – REAL (KIND=nag_wp) arrayOutput
On exit: let qi denote the number of coefficients (of each type) produced by the wavelet transform at level i, for i=nl,nl-1,,1. These values are returned in DWTLEV. Setting k1=qnl and kj+1=kj+qnl-j+1, for j=1,2,,nl, the coefficients are stored as follows:
Ci, for i=1,2,,k1
Contains the level nl approximation coefficients, anl.
Ci, for i=k1+1,,k2 
Contains the level nl detail coefficients dnl.
Ci, for i=kj+1,,kj+1
Contains the level nl-j+1 detail coefficients, for j=2,3,,nl.
5:     NWL – INTEGERInput
On entry: the number of levels, nl, in the multi-level resolution to be performed.
Constraint: 1NWLlmax, where lmax is the value returned in NWL (the maximum number of levels) by the call to the initialization routine C09AAF.
6:     DWTLEV(NWL+1) – INTEGER arrayOutput
On exit: the number of transform coefficients at each level. DWTLEV1 and DWTLEV2 contain the number, qnl, of approximation and detail coefficients respectively, for the final level of resolution (these are equal); DWTLEVi contains the number of detail coefficients, qnl-i+2, for the (nl-i+2)th level, for i=3,4,,nl+1.
7:     ICOMM(100) – INTEGER arrayCommunication Array
On entry: contains details of the discrete wavelet transform and the problem dimension as setup in the call to the initialization routine C09AAF.
On exit: contains additional information on the computed transform.
8:     IFAIL – INTEGERInput/Output
On entry: IFAIL must be set to 0, -1​ or ​1. If you are unfamiliar with this parameter you should refer to Section 3.3 in the Essential Introduction for details.
For environments where it might be inappropriate to halt program execution when an error is detected, the value -1​ or ​1 is recommended. If the output of error messages is undesirable, then the value 1 is recommended. Otherwise, if you are not familiar with this parameter, the recommended value is 0. When the value -1​ or ​1 is used it is essential to test the value of IFAIL on exit.
On exit: IFAIL=0 unless the routine detects an error or a warning has been flagged (see Section 6).

6  Error Indicators and Warnings

If on entry IFAIL=0 or -1, explanatory error messages are output on the current error message unit (as defined by X04AAF).
Errors or warnings detected by the routine:
On entry, N is inconsistent with the value passed to the initialization routine C09AAF.
On entry, LENC<nc*, where nc* is the number of coefficients that will be generated given the chosen value of NWL.
On entry,NWL<1,
orNWL>lmax, where lmax is the value returned in NWL by the call to the initialization routine C09AAF.
On entry, the initialization routine C09AAF has not been called first or it has been called with WTRANS='S', or the communication array ICOMM has become corrupted.

7  Accuracy

The accuracy of the wavelet transform depends only on the floating point operations used in the convolution and downsampling and should thus be close to machine precision.

8  Further Comments

The wavelet coefficients at each level can be extracted from the output array C using the information contained in DWTLEV on exit (see the descriptions of C and DWTLEV in Section 5). For example, given an input data set, x, denoising can be carried out by applying a thresholding operation to the detail coefficients at every level. The elements Ck1+1:knl+1, as described in Section 5, contain the detail coefficients, d^ij, for i=nl,nl-1,,1 and j=1,2,,qi, where d^ij=dij+σεij and σεij is the transformed noise term. If some threshold parameter α is chosen, a simple hard thresholding rule can be applied as
d- ij = 0, if ​ d^ij α d^ij , if ​ d^ij > α,
taking d-ij to be an approximation to the required detail coefficient without noise, dij. The resulting coefficients can then be used as input to C09CDF in order to reconstruct the denoised signal.
See the references given in the introduction to this chapter for a more complete account of wavelet denoising and other applications.

9  Example

This example performs a multi-level resolution of a dataset using the Daubechies wavelet (see WAVNAM='DB4' in C09AAF) using zero end extensions, the number of levels of resolution, the number of coefficients in each level and the coefficients themselves are reused. The original dataset is then reconstructed using C09CDF.

9.1  Program Text

Program Text (c09ccfe.f90)

9.2  Program Data

Program Data (c09ccfe.d)

9.3  Program Results

Program Results (c09ccfe.r)

C09CCF (PDF version)
C09 Chapter Contents
C09 Chapter Introduction
NAG Library Manual

© The Numerical Algorithms Group Ltd, Oxford, UK. 2012