ENTROPY CODING WITH APPLICATION TO LOSSLESS COMPRESSION OF SEISMIC TRACE HEADERS

摘要

Trace headers are an integral part of seismic data. Uncompressed trace headers become a significant portion of the total volume of seismic data in case the associated trace data samples are compressed. We present an efficient lossless compression algorithm applicable for the trace headers of the SEG-Y standard. The proposed technique uses differential coding and run-length coding in conjunction with class-wise entropy coding of the symbols (run) and the counts (length). Four entropy coding methods are compared: non-conditional semi-adaptive Huffman coding, non-conditional adaptive Huffman coding, non-conditional semi-adaptive arithmetic coding, and finally, conditional adaptive arithmetic coding. Typical compression ratios, in the case of trace headers of the common offset gather examples, are about 27.5:1 for the three non-conditional entropy coding methods and around 250:1 for the conditional entropy coding method. For low and medium orders, the compression ratio increases monotonically as a function of the order of the conditional probability model. The best suited order is 8. After compression, the size of trace headers becomes, independently of the applied entropy coding method, negligible compared to the size of the associated trace data samples.