boofcv.alg.transform.fft

Class GeneralPurposeFFT_F32_2D

java.lang.Object

boofcv.alg.transform.fft.GeneralPurposeFFT_F32_2D

public class GeneralPurposeFFT_F32_2D
extends Object

Computes 2D Discrete Fourier Transform (DFT) of complex and real, float precision data. The size of the data can be an arbitrary number. The code originally comes from General Purpose FFT Package written by Takuya Ooura (http://www.kurims.kyoto-u.ac.jp/~ooura/fft.html). See below for the full history.

This code has a bit of a history. Originally from General Purpose FFT. Which was then ported into JFFTPack written by Baoshe Zhang (http://jfftpack.sourceforge.net/), and then into JTransforms by Piotr Wendykier. The major modification from JTransforms is that the SMP code has been stripped out. It might be added back in once an SMP strategy has been finalized in BoofCV.

Code License: The original license of General Purpose FFT Package is shown below. This file will fall under the same license:

 Copyright Takuya OOURA, 1996-2001

 You may use, copy, modify and distribute this code for any purpose (include commercial use) and without fee.
 Please refer to this package when you modify this code.
 

Author:

Piotr Wendykier (piotr.wendykier@gmail.com), Peter Abeles

Constructor Summary

Constructors

Constructor and Description
GeneralPurposeFFT_F32_2D(int rows, int columns) Creates new instance of DoubleFFT_2D.

Method Summary

Methods

Modifier and Type	Method and Description
void	complexForward(float[] a) Computes 2D forward DFT of complex data leaving the result in a.
void	complexInverse(float[] a, boolean scale) Computes 2D inverse DFT of complex data leaving the result in a.
void	realForward(float[] a) Computes 2D forward DFT of real data leaving the result in a .
void	realForwardFull(float[] a) Computes 2D forward DFT of real data leaving the result in a .
void	realInverse(float[] a, boolean scale) Computes 2D inverse DFT of real data leaving the result in a .
void	realInverseFull(float[] a, boolean scale) Computes 2D inverse DFT of real data leaving the result in a .

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

GeneralPurposeFFT_F32_2D

public GeneralPurposeFFT_F32_2D(int rows,
                        int columns)

Creates new instance of DoubleFFT_2D.

Parameters:

rows - number of rows

columns - number of columns

Method Detail

complexForward

public void complexForward(float[] a)

Computes 2D forward DFT of complex data leaving the result in a. The data is stored in 1D array in row-major order. Complex number is stored as two float values in sequence: the real and imaginary part, i.e. the input array must be of size rows*2*columns. The physical layout of the input data has to be as follows:

 a[k1*2*columns+2*k2] = Re[k1][k2],
 a[k1*2*columns+2*k2+1] = Im[k1][k2], 0<=k1<rows, 0<=k2<columns,
 

Parameters:

a - data to transform

complexInverse

public void complexInverse(float[] a,
                  boolean scale)

Computes 2D inverse DFT of complex data leaving the result in a. The data is stored in 1D array in row-major order. Complex number is stored as two float values in sequence: the real and imaginary part, i.e. the input array must be of size rows*2*columns. The physical layout of the input data has to be as follows:

 a[k1*2*columns+2*k2] = Re[k1][k2],
 a[k1*2*columns+2*k2+1] = Im[k1][k2], 0<=k1<rows, 0<=k2<columns,
 

Parameters:

a - data to transform

scale - if true then scaling is performed

realForward

public void realForward(float[] a)

Computes 2D forward DFT of real data leaving the result in a . This method only works when the sizes of both dimensions are power-of-two numbers. The physical layout of the output data is as follows:

 a[k1*columns+2*k2] = Re[k1][k2] = Re[rows-k1][columns-k2],
 a[k1*columns+2*k2+1] = Im[k1][k2] = -Im[rows-k1][columns-k2],
       0<k1<rows, 0<k2<columns/2,
 a[2*k2] = Re[0][k2] = Re[0][columns-k2],
 a[2*k2+1] = Im[0][k2] = -Im[0][columns-k2],
       0<k2<columns/2,
 a[k1*columns] = Re[k1][0] = Re[rows-k1][0],
 a[k1*columns+1] = Im[k1][0] = -Im[rows-k1][0],
 a[(rows-k1)*columns+1] = Re[k1][columns/2] = Re[rows-k1][columns/2],
 a[(rows-k1)*columns] = -Im[k1][columns/2] = Im[rows-k1][columns/2],
       0<k1<rows/2,
 a[0] = Re[0][0],
 a[1] = Re[0][columns/2],
 a[(rows/2)*columns] = Re[rows/2][0],
 a[(rows/2)*columns+1] = Re[rows/2][columns/2]
 

This method computes only half of the elements of the real transform. The other half satisfies the symmetry condition. If you want the full real forward transform, use realForwardFull. To get back the original data, use realInverse on the output of this method.

Parameters:

a - data to transform

realForwardFull

public void realForwardFull(float[] a)

Computes 2D forward DFT of real data leaving the result in a . This method computes full real forward transform, i.e. you will get the same result as from complexForward called with all imaginary part equal 0. Because the result is stored in a, the input array must be of size rows*2*columns, with only the first rows*columns elements filled with real data. To get back the original data, use complexInverse on the output of this method.

Parameters:

a - data to transform

realInverse

public void realInverse(float[] a,
               boolean scale)

Computes 2D inverse DFT of real data leaving the result in a . This method only works when the sizes of both dimensions are power-of-two numbers. The physical layout of the input data has to be as follows:

 a[k1*columns+2*k2] = Re[k1][k2] = Re[rows-k1][columns-k2],
 a[k1*columns+2*k2+1] = Im[k1][k2] = -Im[rows-k1][columns-k2],
       0<k1<rows, 0<k2<columns/2,
 a[2*k2] = Re[0][k2] = Re[0][columns-k2],
 a[2*k2+1] = Im[0][k2] = -Im[0][columns-k2],
       0<k2<columns/2,
 a[k1*columns] = Re[k1][0] = Re[rows-k1][0],
 a[k1*columns+1] = Im[k1][0] = -Im[rows-k1][0],
 a[(rows-k1)*columns+1] = Re[k1][columns/2] = Re[rows-k1][columns/2],
 a[(rows-k1)*columns] = -Im[k1][columns/2] = Im[rows-k1][columns/2],
       0<k1<rows/2,
 a[0] = Re[0][0],
 a[1] = Re[0][columns/2],
 a[(rows/2)*columns] = Re[rows/2][0],
 a[(rows/2)*columns+1] = Re[rows/2][columns/2]
 

This method computes only half of the elements of the real transform. The other half satisfies the symmetry condition. If you want the full real inverse transform, use realInverseFull.

Parameters:

a - data to transform

scale - if true then scaling is performed

realInverseFull

public void realInverseFull(float[] a,
                   boolean scale)

Computes 2D inverse DFT of real data leaving the result in a . This method computes full real inverse transform, i.e. you will get the same result as from complexInverse called with all imaginary part equal 0. Because the result is stored in a, the input array must be of size rows*2*columns, with only the first rows*columns elements filled with real data.

Parameters:

a - data to transform

scale - if true then scaling is performed