DCTInv

Syntax

Parameters

pDCTSpec	Pointer to the inverse DCT specification structure.
pSrc	Pointer to the source vector.
pDst	Pointer to the destination vector.
pSrcDst	Pointer to the source and destination vector for in-place operations.
pBuffer	Pointer to the external work buffer, may be NULL.
scaleFactor	Scale factor, refer to Integer Scaling.

Description

The function ippsDCTInv is declared in the ipps.h file. This function computes the inverse discrete cosine transform (DCT) of the source signal pSrc (pSrcDst for in-place operations) in accordance with the specification structure pDCTSpec that must be initialized by calling the functions ippsDCTInvInitAlloc or ippsDCTInvInit beforehand. The result in stored in the pDst (pSrcDst for in-place operations).

If len is a power of 2, the functions use an efficient algorithm that is significantly faster than the direct computation of DCT. For other values of len, these functions use the direct formulas given below; however, the symmetry of the cosine function is taken into account, which allows to perform about half of the multiplication operations in the formulas.

In the following definition of DCT, N = len,

x(n) is pDst[n] and y(k) is pSrc[k].

The inverse DCT is defined by the formula:

For integer data types the output result is scaled according to the scaleFactor value, thus the output signal range and precision are retained.

The function may be used with the external work buffer pBuffer to avoid memory allocation within the functions. Once the work buffer is allocated, it can be used for all following calls to the functions computing DCT. As internal allocation of memory is too expensive operation and depends on operating system and/or runtime libraries used - the use of an external buffer improves performance significantly, especially for the small size transforms.

The size of this buffer must be computed previously using the functions ippsDCTInvGetBufSize or ippsDCTInvGetSize.

If the external buffer is not specified (pBuffer is set to NULL), then the function itself allocates the memory needed for operation.

Example below shows how to use the functions ippsDCTFwd_32f and ippsDCTInv_32f to compress and reconstruct a signal.

Return Values

ippStsNoErr	Indicates no error.
ippStsNullPtrErr	Indicates an error if one of the pDCTSpec, pSrc, pDst, pSrcDst pointers is NULL.
ippStsContextMatchErr	Indicates an error when the specification identifier pDCTSpec is incorrect.
ippStsMemAllocErr	Indicates an error if memory allocation fails.

Using ippsDCTFwd and ippsDCTInv to Compress and Reconstruct a Signal

void dct( void ) {

#define LEN 256

     Ipp32f x[LEN], y[LEN];

     int n;

     IppsDCTFwdSpec_32f* fspec;

     IppsDCTInvSpec_32f* ispec;

     IppStatus status;

     /// data: Gaussian function, magn =1 and sigma=N/3 

     for(n=0; n<LEN; ++n) 

         x[n] = (float)(exp(-0.5*(LEN/2-n)*(LEN/2-n)/(LEN/3.0)));

     /// get cosine transform coefficients

     status = ippsDCTFwdInitAlloc_32f( &fspec, LEN, ippAlgHintNone );

     status = ippsDCTFwd_32f( x, y, fspec, NULL );

     ippsDCTFwdFree_32f( fspec );

     /// Set 3/4 of these coefficients to zero

     for(n=LEN/4; n<LEN; ++n) y[n] = 0.0f;

     /// restore signal using len/4 values

     status = ippsDCTInvInitAlloc_32f( &ispec, LEN, ippAlgHintNone );

     status = ippsDCTInv_32f( y, x, ispec, NULL );

     ippsDCTInvFree_32f( ispec );

}