Adaptive seismic compression by wavelet shrinkage

Abstract

In this paper, a sophisticated adaptive seismic compression method is presented based on wavelet shrinkage. Our approach combines a time-scale transform with an adaptive non-linear statistical method. First, a discrete 2-D biorthogonal Discrete Wavelet Transform (DWT) is applied to the multi-channel seismic signals to generate a sparse multiresolution (subband) decomposition. Compression is then achieved by shrinking the detail wavelet coefficients using a scale-dependent non-linear soft-thresholding rule. The adaptive scale-dependent thresholds are determined by minimizing the Stein's Unbiased Risk Estimate (SURE). The proposed compression procedure is tested on marine seismic data from the Midyan basin (Red Sea, Saudi Arabia).

Statistical Atmospheric Parameter Retrieval Largely Benefits From Spatial–Spectral Image Compression

“…There is a wide evidence that compressing the signal before any further processing is beneficial. We have shown this effect previously for several compression algorithms in hyperspectral image classification [48], and it has been a common strategy in other applications such as genomics -in particular for Next Generation Sequencing (NGS) [49], [50]-, seismic signal processing [51], [52], bioengineering [53] and communications [54]. Therefore, we posit that what we report is yet another case of an effect previously noticed in signal and image processing, here transported to the field of bio-geo-physical parameter estimation in atmospheric applications using infrared sounding data.…”

Adaptive Seismic Data Compression Using Wavelet Packets

“…Some of the algorithms used up to now for seismic data compression were based on some type of transforms, such as local cosine transform, or wavelet transform, while combining with a uniform quantization and. Huffman coding schemes [1][2][3][4][5]. Using these family of algorithms may achieve low to moderate compression rations, while seeking higher compression rations yield significant artifacts in the reconstructed data.…”