Wavelets are ill-suited to represent oscillatory patterns: rapid variations
of intensity can only be described by the small scale wavelet coefficients
, which are often quantized to zero, even at high bit rates. Our goal in th
is paper is to provide a fast numerical implementation of the best wavelet
packet algorithm [1] in order to demonstrate that an advantage can be gaine
d by constructing a basis adapted to a target image. Emphasis in this paper
has been placed on developing algorithms that are computationally efficien
t. We developed a new fast two-dimensional (2-D) convolution-decimation alg
orithm with factorized nonseparable 2-D filters. The algorithm is four time
s faster than a standard convolution-decimation, An extensive evaluation of
the algorithm was performed on a large class of textured images. Because o
f its ability to reproduce textures so well, the wavelet packet coder signi
ficantly out performs one of the best wavelet coder [2] on images such as B
arbara and fingerprints, both visually and in term of PSNR.