Efficient Breadth-First Implementation of the Wavelet Transform

Cache-efficient breadth-first lifting-based algorithms for the wavelet transform (WT) are considered. Two optimizations for improving the efficiency of the WT computation are proposed. These optimizations are then applied to two different baseline WT algorithms and shown to be highly effective, reducing the execution time by as much as one third. Moreover, the resulting optimized WT algorithms are shown to be quite competitive with another more sophisticated algorithm, in spite of having very substantially reduced memory requirements