NAG Library Routine Document

c09fcf computes the three-dimensional multi-level discrete wavelet transform (DWT). The initialization routine c09acf must be called first to set up the DWT options.

2

Specification

Fortran Interface

Subroutine c09fcf (

m, n, fr, a, lda, sda, lenc, c, nwl, dwtlvm, dwtlvn, dwtlvfr, icomm, ifail)

Integer, Intent (In)	::	m, n, fr, lda, sda, lenc, nwl
Integer, Intent (Inout)	::	icomm(260), ifail
Integer, Intent (Out)	::	dwtlvm(nwl), dwtlvn(nwl), dwtlvfr(nwl)
Real (Kind=nag_wp), Intent (In)	::	a(lda,sda,fr)
Real (Kind=nag_wp), Intent (Inout)	::	c(lenc)

C Header Interface

#include <nagmk26.h>

void	c09fcf_ (const Integer m, const Integer n, const Integer fr, const double a[], const Integer lda, const Integer sda, const Integer lenc, double c[], const Integer nwl, Integer dwtlvm[], Integer dwtlvn[], Integer dwtlvfr[], Integer icomm[], Integer ifail)

3

Description

c09fcf computes the multi-level DWT of three-dimensional data. For a given wavelet and end extension method, c09fcf will compute a multi-level transform of a three-dimensional array

A

, using a specified number,

n_{fwd}

, of levels. The number of levels specified,

n_{fwd}

, must be no more than the value

l_{\max}

returned in nwlmax by the initialization routine c09acf 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 data,

A

. Level 1 consists of the first set of coefficients computed: the seven sets of detail coefficients are stored at this level while the approximation coefficients are used as the input to a repeat of the wavelet transform at the next level. This process is continued until, at level

n_{fwd}

, all eight types of coefficients are stored. All coefficients are packed into a single array.

4

References

Wang Y, Che X and Ma S (2012) Nonlinear filtering based on 3D wavelet transform for MRI denoising URASIP Journal on Advances in Signal Processing 2012:40

5

Arguments

1: $m$ – IntegerInput

On entry: the number of rows of each two-dimensional frame.

Constraint: this must be the same as the value m passed to the initialization routine c09acf.

2: $n$ – IntegerInput

On entry: the number of columns of each two-dimensional frame.

Constraint: this must be the same as the value n passed to the initialization routine c09acf.

3: $fr$ – IntegerInput

On entry: the number of two-dimensional frames.

Constraint: this must be the same as the value fr passed to the initialization routine c09acf.

4: $a (lda, sda, fr)$ – Real (Kind=nag_wp) arrayInput

On entry: the

m

n

fr

three-dimensional input data

A

, where with

A_{i j k}

stored in

a (i, j, k)

5: $lda$ – IntegerInput

On entry: the first dimension of the array a as declared in the (sub)program from which c09fcf is called.

Constraint:

lda \geq m

6: $sda$ – IntegerInput

On entry: the second dimension of the array a as declared in the (sub)program from which c09fcf is called.

Constraint:

sda \geq n

7: $lenc$ – IntegerInput

On entry: the dimension of the array c as declared in the (sub)program from which c09fcf is called.

Constraint:

lenc \geq n_{ct}

, where

n_{ct}

is the total number of wavelet coefficients that correspond to a transform with nwl levels.

8: $c (lenc)$ – Real (Kind=nag_wp) arrayOutput

On exit: the coefficients of the discrete wavelet transform. If you need to access or modify the approximation coefficients or any specific set of detail coefficients then the use of c09fyf or c09fzf is recommended. For completeness the following description provides details of precisely how the coefficients are stored in c but this information should only be required in rare cases.

Let

q (i)

denote the number of coefficients of each type at level

i

, for

i = 1, 2, \dots, n_{fwd}

, such that

q (i) = dwtlvm (n_{fwd} - i + 1) \times dwtlvn (n_{fwd} - i + 1) \times dwtlvfr (n_{fwd} - i + 1)

. Then, letting

k_{1} = q (n_{fwd})

and

k_{j + 1} = k_{j} + q (n_{fwd} - ⌈j / 7⌉ + 1)

, for

j = 1, 2, \dots, 7 n_{fwd}

, the coefficients are stored in c as follows:

$c (i)$ , for $i = 1, 2, \dots, k_{1}$

Contains the level

n_{fwd}

approximation coefficients,

a_{n_{fwd}}

. Note that for computational efficiency reasons these coefficients are stored as

dwtlvm (1) \times dwtlvn (1) \times dwtlvfr (1)

in c.

$c (i)$ , for $i = k_{j} + 1, \dots, k_{j + 1}$

Contains the level

n_{fwd} - ⌈j / 7⌉ + 1

detail coefficients. These are:

LLH coefficients if $j mod 7 = 1$ ;
LHL coefficients if $j mod 7 = 2$ ;
LHH coefficients if $j mod 7 = 3$ ;
HLL coefficients if $j mod 7 = 4$ ;
HLH coefficients if $j mod 7 = 5$ ;
HHL coefficients if $j mod 7 = 6$ ;
HHH coefficients if $j mod 7 = 0$ ,

for

j = 1, \dots, 7 n_{fwd}

. See Section 2.1 in the C09 Chapter Introduction for a description of how these coefficients are produced.

Note that for computational efficiency reasons these coefficients are stored as

dwtlvfr (⌈j / 7⌉) \times dwtlvm (⌈j / 7⌉) \times dwtlvn (⌈j / 7⌉)

in c.

9: $nwl$ – IntegerInput

On entry: the number of levels,

n_{fwd}

, in the multi-level resolution to be performed.

Constraint:

1 \leq nwl \leq l_{\max}

, where

l_{\max}

is the value returned in nwlmax (the maximum number of levels) by the call to the initialization routine c09acf.

10: $dwtlvm (nwl)$ – Integer arrayOutput

On exit: the number of coefficients in the first dimension for each coefficient type at each level.

dwtlvm (i)

contains the number of coefficients in the first dimension (for each coefficient type computed) at the (

n_{fwd} - i + 1

)th level of resolution, for

i = 1, 2, \dots, n_{fwd}

11: $dwtlvn (nwl)$ – Integer arrayOutput

On exit: the number of coefficients in the second dimension for each coefficient type at each level.

dwtlvn (i)

contains the number of coefficients in the second dimension (for each coefficient type computed) at the (

n_{fwd} - i + 1

)th level of resolution, for

i = 1, 2, \dots, n_{fwd}

12: $dwtlvfr (nwl)$ – Integer arrayOutput

On exit: the number of coefficients in the third dimension for each coefficient type at each level.

dwtlvfr (i)

contains the number of coefficients in the third dimension (for each coefficient type computed) at the (

n_{fwd} - i + 1

)th level of resolution, for

i = 1, 2, \dots, n_{fwd}

13: $icomm (260)$ – 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 c09acf.

On exit: contains additional information on the computed transform.

14: $ifail$ – IntegerInput/Output

On entry: ifail must be set to

0

- 1 or 1

. If you are unfamiliar with this argument you should refer to Section 3.4 in How to Use the NAG Library and its Documentation 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 argument, 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

- 1

, explanatory error messages are output on the current error message unit (as defined by x04aaf).

Errors or warnings detected by the routine:

$ifail = 1$: On entry, $fr = 〈value〉$ .
Constraint: $fr = 〈value〉$ , the value of fr on initialization (see c09acf).

On entry, $m = 〈value〉$ .
Constraint: $m = 〈value〉$ , the value of m on initialization (see c09acf).

On entry, $n = 〈value〉$ .
Constraint: $n = 〈value〉$ , the value of n on initialization (see c09acf).

$ifail = 2$: On entry, $lda = 〈value〉$ and $m = 〈value〉$ .
Constraint: $lda \geq m$ .

On entry, $sda = 〈value〉$ and $n = 〈value〉$ .
Constraint: $sda \geq n$ .

$ifail = 3$: On entry, $lenc = 〈value〉$ .
Constraint: $lenc \geq 〈value〉$ , the total number of coefficents to be generated.

$ifail = 5$: On entry, $nwl = 〈value〉$ .
Constraint: $nwl \geq 1$ .

On entry, $nwl = 〈value〉$ and $nwlmax = 〈value〉$ in c09acf.
Constraint: $nwl \leq nwlmax$ in c09acf.

$ifail = 6$: Either the communication array icomm has been corrupted or there has not been a prior call to the initialization routine c09acf.

The initialization routine was called with $wtrans ='S'$ .

$ifail = - 99$: An unexpected error has been triggered by this routine. Please contact NAG.
See Section 3.9 in How to Use the NAG Library and its Documentation for further information.

$ifail = - 399$: Your licence key may have expired or may not have been installed correctly.
See Section 3.8 in How to Use the NAG Library and its Documentation for further information.

$ifail = - 999$: Dynamic memory allocation failed.
See Section 3.7 in How to Use the NAG Library and its Documentation for further information.

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

Parallelism and Performance

c09fcf is not threaded in any implementation.

9

Further Comments

The example program shows how the wavelet coefficients at each level can be extracted from the output array c. Denoising can be carried out by applying a thresholding operation to the detail coefficients at every level. If

c_{i j}

is a detail coefficient then

{\hat{c}}_{i j} = c_{i j} + σ ε_{i j}

and

σ ε_{i j}

is the transformed noise term. If some threshold parameter

α

is chosen, a simple hard thresholding rule can be applied as

{\bar{c}}_{i j} = \{\begin{cases} 0, & if ​ |{\hat{c}}_{i j}| \leq α \\ {\hat{c}}_{i j}, & if ​ |{\hat{c}}_{i j}| > α, \end{cases}

taking

{\bar{c}}_{i j}

to be an approximation to the required detail coefficient without noise,

c_{i j}

. The resulting coefficients can then be used as input to c09fdf in order to reconstruct the denoised signal. See Section 10 in c09fzf for a simple example of denoising.

See the references given in the introduction to this chapter for a more complete account of wavelet denoising and other applications.

10

Example

This example computes the three-dimensional multi-level discrete wavelet decomposition for

7 \times 6 \times 5

input data using the biorthogonal wavelet of order

1.1

(set

wavnam ='BIOR1.1'

in c09acf) with periodic end extension, prints a selected set of wavelet coefficients and then reconstructs and verifies that the reconstruction matches the original data.

NAG Library Routine Document

c09fcf (dim3_multi_fwd)

▸▿ Contents

1 Purpose

2 Specification

3 Description

4 References

5 Arguments

6 Error Indicators and Warnings

7 Accuracy

8 Parallelism and Performance

9 Further Comments

10 Example

10.1 Program Text

10.2 Program Data

10.3 Program Results

1

Purpose

2

Specification

3

Description

4

References

5

Arguments

6

Error Indicators and Warnings

7

Accuracy

8

Parallelism and Performance

9

Further Comments

10

Example

10.1

Program Text

10.2

Program Data

10.3

Program Results