| A vector-valued wavelet (multlwavelet) is a function Y=y1 y2 with yj∈L2R and such that the collection 2n2y j2nx-k;n, k∈Z,j=1,&ldots;,N is an orthonormal basis for L 2(R). A multiwavelet-based compression works with vector valued data. The standard approach is first to pre-process scalar data into two channels. In this work we explore the use of multiwavelets to compress stereo signals.; We implement a compression scheme based on multi-wavelet transform and test performance of standard multiwavelets, compared to traditional (scalar) wavelet transform. It turns out that the standard multiwavelets do not have an advantage over compressing each channel separately. The properties that allow good compression of stereo data are different from the properties of good scalar wavelets. We construct multiwavelets with various combinations of desirable properties, using a general method we developed. In some cases, the compression results are indeed better than those achieved by any standard method. |
