GPU快速解压的多分辨率地形高程数据压缩算法

Multi-resolution terrain height-field data compression for rapid GPU decoding

为了缓解实时地形可视化中海量数据的存储和传输压力,在构建地形数据多分辨率金字塔模型的基础上,提出了一种适于GPU实时解压的分块高程数据无损压缩和有损压缩算法。该算法首先对地形块数据进行线性预测,对高分辨率地形块的预测误差采用递归变长的无损编码方法,保留了高分辨率地形数据的细节特征;对低分辨率地形块的预测误差,在考虑地形块本身简化误差的基础上采用了分组的有损量化编码方法。给出了GPU上实现快速并行解压的处理流程。实验结果表明,该算法实现了有效的数据压缩,具有很高的实时解压效率。

To release the burden for massive-data storage and transmission in real-time large terrain visualization systems, based on the multi-resolution pyramid model for terrain data, this paper proposed a compression algorithm for terrain blocks that was suitable for fast decoding＇on the GPU. At first, the algorithm implemented linear prediction for terrain blocks, and then coded residual errors for lossless compression of the full-resolution data based on RBUC （recursive bottom up complete）. The lossless residual coding method reserved the details of the full-resolution data. As for the residual errors on coarser levels in the terrain blocks, it adopted a grouped and quantized coding method for lossy compression. The quantization levels were based on the simplified errors of low-resolution blocks. At last, it designed the processing flow for fast and parallel decompression on GPU. The experimental results show that the algorithm can be used to compress terrain height-field datasets effectively and supports real-time decompression with high efficiency.