A Parallel Algorithm for Multi-Resolution Representation of DEM Based on Discrete Wavelet Analysis

Because of the discrete wavelets characteristics of multi-resolution, discrete wavelet transformation can be used to construct multi-resolution of the DEM data. Nevertheless, the intensive computation of DWT has become a significant bottleneck in real-time applications when the size of DEM data is large enough. It presents a parallel processing framework to overcome the drawbacks of this problem and achieves a fast parallel building of the DEM datas multi-resolution model. As the DEM data is equivalent to two-dimensional gray-scale images, it can be processed as two-dimensional signal. This paper first gives the Mallat algorithm for a two-dimensional discrete wavelet transform (DWT). Based on this algorithm, the parallel property of the DWT is analyzed. Result shows that the Mallat algorithm is easy to carry out in parallel mode because of the local nature of the DWT. Then analyzed are the data communications between the logical topology of the multi-processors. Concretely, the authors present a data structure of the local array declared on the processor and describe the communication relationships between the various processors on two-dimensional grids. Simulation tests show that the parallel model leads to a satisfactory scalability of the algorithm and high speedups are achieved.