Transforms » Low-level transform support module
#include "diplib/dft.h"

Low-level functionality for computing the Discrete Fourier Transform.

Contents

Classes

template<typename T>
class dip::DFT
An object that encapsulates the Discrete Fourier Transform (DFT).
template<typename T>
class dip::RDFT
An object that encapsulates the real-valued Discrete Fourier Transform (DFT).

Aliases

using dip::Option::DFTOptions = dip::detail::Options
Determines working mode for dip::DFT and dip::RDFT. Combines multiple dip::Option::DFTOption values,.

Enums

enum class dip::Option::DFTOption: uint8{ InPlace, TrashInput, Aligned }
Determines working mode for dip::DFT and dip::RDFT.

Functions

auto dip::GetOptimalDFTSize(dip::uint size0, bool larger = true, dip::uint maxFactor = 5) -> dip::uint
Returns a size equal or larger to size0 that is efficient for the DFT implementation. The value returned is a product of small primes.
auto dip::MaxFactor(bool complex = true) -> dip::uint
The maxFactor parameter for dip::GetOptimalDFTSize. complex determines whether the transform to be computed is complex-to-complex or not.

Variables

dip::uint const dip::maximumDFTSize
The largest size supported by dip::DFT and dip::FourierTransform. Is equal to 231-1 when using FFTW, 264-1 when using PocketFFT on a 64-bit system, or 232-1 when using PocketFFT on a 32-bit system.
bool const dip::usingFFTW
Is true if dip::DFT and dip::RDFT use the FFTW library, or false if they use PocketFFT.

Enum documentation

enum class dip::Option::DFTOption: uint8

Determines working mode for dip::DFT and dip::RDFT.

Implicitly casts to dip::Option::DFTOptions. Combine constants together with the + operator.

Enumerators
InPlace = 0 Work in place, the input and output buffers are the same.
TrashInput = 1 Allowed to trash the input buffer, we don’t need to preserve it.
Aligned = 2 Both buffers are aligned to 16-byte boundaries.

Function documentation

dip::uint dip::GetOptimalDFTSize(dip::uint size0, bool larger = true, dip::uint maxFactor = 5)

Returns a size equal or larger to size0 that is efficient for the DFT implementation. The value returned is a product of small primes.

Set larger to false to return a size equal or smaller instead.

maxFactor should be 5, 7 or 11, and gives the largest integer that should be considered a “small prime”. When using FFTW, set maxFactor to 7, as by default it works most efficiently for sizes that can be factored into primes smaller or equal to 7. When using PocketFFT, set it to 5 for complex-to-real or real-to-complex transforms, and to 11 for complex-to-complex transforms. See dip::MaxFactor.

If there is no suitable size that the implementation can use (see dip::maximumDFTSize), this function will return 0.

Prefer to use dip::OptimalFourierTransformSize in your applications, it will throw an error if the transform size is too large, and determines maxFactor for you.