The Research Of Seismic Data Compression Based On Wavelet Transform

With the developing of seismic exploration technique in the direction of multidimensional, high-resolution, multi-parameter and multi-component, the seismic data is exponentially increasing, which causes great pressure and challenges to the conventional memory capacity, processing speed of computer, network bandwidth required for the transmission of data. Therefore, it is very significant to reduce redundant information by data compression. According to the characteristics of 2D seismic data, this paper has selected two methods, JPEG2000 core algorithm and EZW algorithm, which are both based on wavelet transform. Pre-stack seismic data and post-stack seismic data will be compressed with these two methods respectively.In the experiment of data compression with JPEG2000 core algorithm, a modified JPEG2000 system framework is used to process the seismic data with amplitude of high dynamic range. Firstly, a five-level W5/3 wavelet transform with lifting scheme is used to process the original data, by this operation, the most energy of original data is concentrated on a few coefficients of lower frequency subbands; then some redundant information will be eliminated by quantization method; then quantized coefficients will be sent into Tier-1 encoder, using MQ arithmetic encoder to encode the coefficients into bit stream through predicting the imporatance of coefficents by using 8-neighborhood coefficients context; lastly, in Tier-2 encoder, some packaged data units will be threw off according to the compression ratio. Experimental results show that the signal-to-noise ratio (SNR) can reach 40dB or more when pre-stack seismic data is compressed at a compression ratio of 5:1, while the SNR can reach 30dB or more when post-stack seismic data is compressed at a compression ratio of 15:1. With above efforts, JPEG2000 core algorithm can be successfully appled into seismic data compression.In the experiment of data compression with EZW algorithm, a modified EZW system framework is used to process the seismic data with amplitude of high dynamic range. Firstly, a five-level wavelet transform with Mallat fast decomposition algorithm is used to process the original data, by this operation, the most energy of original data is concentrated on a few coefficients of lower frequency subbands; then some redundant information will be eliminated by quantization method; then quantized coefficients will be sent into the EZW encoder, here all coefficients will calarified into four kinds:POS、 NEG、ZTR、IZ by the tree-based structure of wavelet transform domain, this step will output the bit sream. In order to further improve the compression times, a adaptive arithmetic encoder will be used to output the final bit stream. Experimental results show that the signal-to-noise ratio (SNR) can be up to 50dB above when pre-stack seismic data is compressed at a compression ratio of 4:1, while the SNR can be up to 30dB above when post-stack seismic data is compressed at a compression ratio of 16:1. With above efforts, EZW algorithm can be successfully appled into seismic data compression.The results of this paper further indicate that the structural characteristics of 2D seismic data are different from common image data. The compression quality is greatly relative to the characteristics of wavelet basis and similarity of the seismic signal waveform and wavelet function.