topic How to form the codebook for vector quantization functions in Intel® Integrated Performance Primitives

How to form the codebook for vector quantization functions

khalil_mi — Tue, 20 Dec 2005 14:59:50 GMT

I try to use the vector quantization functions but I can not form the codebook

Is there any sample code that shows how to use these functions

Thanks

Mahmoud

Note: I use the following package:

Package ID: w_ipp_ia32_itanium_pu_4.1.3.002
Package Contents: Intel Integrated Performance Primitives v4.1 for Windows* on Intel Pentium and Itanium processors

Re: How to form the codebook for vector quantization functions

Vladimir_Dudnik — Wed, 21 Dec 2005 00:54:09 GMT

Hi Mahmoud,

could you please specify which exactly functions you are trying to use?

Regards,
Vladimir

Re: How to form the codebook for vector quantization functions

khalil_mi — Thu, 22 Dec 2005 07:20:17 GMT

Thanks Vladimir

I try to use the ippsVQDistSingle_Sort_16s32s_Sfs function

The vectors I have are 32 integer numbers

I believe I have to use the ippsCdbkInit_L2_16s function first to allocate the codeBook

Mahmoud

Re: How to form the codebook for vector quantization functions

Intel_C_Intel — Fri, 23 Dec 2005 16:08:56 GMT

Mahmoud,

IPP support vector quantisation for 32f and 16sdata. initCdbk_L2 functions allocates memory and builds the codebook. Full and tree search are supported.

ippsVQDistSingleSort_16s32s_Sfs functions finds several best clusters for 16s vectors and calculates distance to centroids as scaled32s numbers.

To do VQ for 32s data with IPP functions you should convert them to 32f or 16s format.

Alexander