Chris@19: Chris@19: Chris@19: 1d Real-even DFTs (DCTs) - FFTW 3.3.4 Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19: Chris@19:
Chris@19: Chris@19: Chris@19:

Chris@19: Next: , Chris@19: Previous: The 1d Real-data DFT, Chris@19: Up: What FFTW Really Computes Chris@19:

Chris@19:
Chris@19: Chris@19:

4.8.3 1d Real-even DFTs (DCTs)

Chris@19: Chris@19:

The Real-even symmetry DFTs in FFTW are exactly equivalent to the unnormalized Chris@19: forward (and backward) DFTs as defined above, where the input array Chris@19: X of length N is purely real and is also even symmetry. In Chris@19: this case, the output array is likewise real and even symmetry. Chris@19: Chris@19: Chris@19:

For the case of REDFT00, this even symmetry means that Chris@19: Xj = XN-j,where we take X to be periodic so that Chris@19: XN = X0. Because of this redundancy, only the first n real numbers are Chris@19: actually stored, where N = 2(n-1). Chris@19: Chris@19:

The proper definition of even symmetry for REDFT10, Chris@19: REDFT01, and REDFT11 transforms is somewhat more intricate Chris@19: because of the shifts by 1/2 of the input and/or output, although Chris@19: the corresponding boundary conditions are given in Real even/odd DFTs (cosine/sine transforms). Because of the even symmetry, however, Chris@19: the sine terms in the DFT all cancel and the remaining cosine terms are Chris@19: written explicitly below. This formulation often leads people to call Chris@19: such a transform a discrete cosine transform (DCT), although it is Chris@19: really just a special case of the DFT. Chris@19: Chris@19: Chris@19:

In each of the definitions below, we transform a real array X of Chris@19: length n to a real array Y of length n: Chris@19: Chris@19:

REDFT00 (DCT-I)
Chris@19: Chris@19:

An REDFT00 transform (type-I DCT) in FFTW is defined by: Chris@19:

.
Note that this transform is not defined for n=1. For n=2, Chris@19: the summation term above is dropped as you might expect. Chris@19: Chris@19:
REDFT10 (DCT-II)
Chris@19: Chris@19:

An REDFT10 transform (type-II DCT, sometimes called “the” DCT) in FFTW is defined by: Chris@19:

.
Chris@19: Chris@19:
REDFT01 (DCT-III)
Chris@19: Chris@19:

An REDFT01 transform (type-III DCT) in FFTW is defined by: Chris@19:

.
In the case of n=1, this reduces to Chris@19: Y0 = X0. Up to a scale factor (see below), this is the inverse of REDFT10 (“the” DCT), and so the REDFT01 (DCT-III) is sometimes called the “IDCT”. Chris@19: Chris@19: Chris@19:
REDFT11 (DCT-IV)
Chris@19: Chris@19:

An REDFT11 transform (type-IV DCT) in FFTW is defined by: Chris@19:

.
Chris@19: Chris@19:
Inverses and Normalization
Chris@19: Chris@19:

These definitions correspond directly to the unnormalized DFTs used Chris@19: elsewhere in FFTW (hence the factors of 2 in front of the Chris@19: summations). The unnormalized inverse of REDFT00 is Chris@19: REDFT00, of REDFT10 is REDFT01 and vice versa, and Chris@19: of REDFT11 is REDFT11. Each unnormalized inverse results Chris@19: in the original array multiplied by N, where N is the Chris@19: logical DFT size. For REDFT00, N=2(n-1) (note that Chris@19: n=1 is not defined); otherwise, N=2n. Chris@19: Chris@19: Chris@19:

In defining the discrete cosine transform, some authors also include Chris@19: additional factors of Chris@19: √2(or its inverse) multiplying selected inputs and/or outputs. This is a Chris@19: mostly cosmetic change that makes the transform orthogonal, but Chris@19: sacrifices the direct equivalence to a symmetric DFT. Chris@19: Chris@19: Chris@19: Chris@19: