Abstract:Through the real cases,the paper discussed the process using balanced biorthogonal multi-wavelets transform to implement noise-elimination and data compression of seismic data.The processes of noise-elimination are follows:①using components of high-frequency and high-wavenumber to evaluate variance of random noise in multi-wavelets domain;②variation of seismic data minus variance of noise is equal to variance of useful signal,then threshold is computed,using soft threshold value processing method for noise-elimination;③the noise-eliminated signals are resulted from the inversion wavelet transform.The processes of data compression are follows:①decomposing the seismic data into 4 layers by balanced biorthogonal wavelets BSA6/6,the decomposition is carried out only for low-frequency parts each time;②rearranging each 2×2 sub-block so that the 4 coefficients that have same special positions are arranged together (original 16 sub-blocks now became 4 subblocks),that met the parent-child relationship required by SHIFT coding scheme;③embedded coding is carried out for the tree of each level,using improved wavelet zero-tree code scheme to compress seismic data.Experiment results showed that the method in the paper is superior to the noise-elimination and data compression methods by single wavelet.