Fast 3-D seismic adaptive compression using the lifted wavelet transform

In the present contribution, a new fast 3-D seismic adaptive compression strategy based on the Lifted (discrete) Wavelet Transform (LWT) is proposed to improve the overall performance of modem 3-D seismic interpretation. The compression scheme consists of three units, namely a 3-D biorthogonal LWT, an adaptive quantizer, and an entropy encoder. The LWT constitutes the most efficient implementation of higher dimension wavelet transforms. It allows for an in-place implementation, which saves considerable auxiliary memory usage. This turns out to be of the utmost importance when compressing large datasets like seismic data. Moreover, the LWT is immediately invertible, with exactly the same complexity as for the forward transform. The proposed adaptive quantizer involves techniques from non-linear approximation theory, namely the wavelet shrinkage. Indeed the scale and direction dependent dead-zone of the Threshold Uniform Scalar Quantizer (TUSQ) is determined for each subband using the SureShrink principle. The entropy encoder is the classical association run-length and Huffman encoding techniques.