The data occupy a volume in four dimensions: time, receiver number, shot number and cable number. Large scale correlations caused by gross trends in subsurface structure are often present along two or more of these dimensions and are well exploited by wavelet compression methods. Further, uncertainty as to the size of a geologic target often leads to local oversampling, allowing further compression with little loss of data integrity.
Compression is carried out by a variety of biorthogonal wavelet filters employed along multi-dimensional transform trees, which may be selected at run time. The filters and transform trees are encoded in the compressed file, providing the flexibility to adapt to the large bandwidth variations found in seismic data. Scalar quantization and entropy coding techniques are used to reduce the wavelet transformed data to a user selected compression ratio. Lossless compression paths involving frame differencing, run length encoding and Huffman coding are used to compress header data which accompanies the seismic data and must be preserved exactly.
Compression ratios of between 20:1 and 100:1 are currently being used to allow transmission of raw and partially processed seismic data from the point of acquisition to land based processing centers. The primary data links are point to point, demand assigned, 64Kbaud, HSD Inmarsat and 128 Kbaud regional leased systems. While seismic data compression is not yet universal in the industry, reductions of between 1 and 2 orders of magnitude in wide area networking costs appear to be driving its adoption.
|
Modified: Jun 25, 1997|
Current events|
Your comments
and inquiries are welcome.