Chapter I.1. Introduction

Chapter I.1. Introduction
Prev	Part I. KIWIWALI : the standalone wavelet library	Next

The purpose of KIWIWALI is to offer quick access to straightforward wavelet transform routines, while hiding most of the complexity of the internals of Kicksey-Winsey. In particular, if requested by the user, wavelet coefficients are transparently re-ordered according to the classical 'non-standard' scheme used also in Wavelab and several other wavelet transform libraries.

Moreover, all necessary templates are meant to be instanciated only once when compiling the KIWIWALI on a given system, which means that compilation of user programs is likely to be much faster than when using the template library.

KIWIWALI supports the following features:

N-dimensional orthogonal wavelet transforms
OpenMP parallelization
arbitrary scalar datatypes
classical 'non-standard'/wavelab coefficient ordering
Kicksey-Winsey interlaced coefficient ordering

The following more advanced features are not supported in the current version of KIWIWALI, but are present in the template wavelet library:

MPI parallelization
dual-tree complex wavelet transforms
biorthogonal wavelet transforms
adaptive 1D wavelet transforms
tensorized/anisotropic wavelet transforms
vector datatypes
runtime filter manipulation
wavelet-based density estimation

Note that, for lack of low level optimization, KIWIWALI is not an extremely fast library, and does not have any claim to compete with, for example, FFTW in terms of efficiency. However, it should scale sufficiently well, and be flexible enough, to be applicable to a wide variety of problems at a reasonnable computational cost.