Summary Report No. 68


A Parallel Algorithm for the Discrete Orthogonal Wavelet Transform

M. Uhlmann (December 2000)


We recall the basics of discrete orthogonal wavelet bases and show how a fast algorithm for the transform of n-dimensional data can be constructed and implemented on distributed memory machines. For this purpose, we use a ’slice’ representation of data across processors and restrict to the case of a power-of-two number of processors for simplicity. Some examples of the transform and filtering of two- and three-dimensional data are given. It is found that our parallel data-model leads to a satisfactory scalability of the algorithm.


Complete document (1.0 MB)