Encoding multidimensional wavelet coefficients using the generalized zerotree

Wavelet transformations are the current state of the art in (lossy) image compression. Efficiently encoding the resulting wavelet coefficients is crucial for the performance of the compression engine, the most commonly used techniques today being the embedded zerotree and SPIHT. Research so far has focussed mostly on image and video compression, resulting in specialized algorithms and data structures for 2D and 3D data. In contrast, a generalized zerotree capable of encoding data of arbitrary dimensionality is presented in this paper, which has been implemented as part of the compression engine of the multidimensional array DBMS RasDaMan. The paper concentrates on some implementation and optimization issues to minimize memory consumption and presents some compression results for 3D and 4D data.