【摘要】 多尺度反演具有对初始模型依赖性低、收敛速度快、反演结果稳定等优点 ,是一种很有希望的反演方法。基于小波变换讨论二维地震波形多尺度反演问题 ,将地震参数反演转化为小波域重要系数的优化问题 ,有效改善了局部极值等。并引入了EZW思想 ,用小波系数的零树来预测待优化参数的位置 ,加速优化参数的搜索 ,使反演更有目的性更多还原

【Abstract】 Multiscale inversion is a promising method because of its low dependence of the initial model, convergence efficiency and robustness. A wavelet multiscale inversion (WMI) method using the orthogonal wavelet transform is applied to inverse two dimension seismic waveform. The inversion result can be re-directed for the optimization of the wavelet significant coefficients (WSC) rather than the traditional seismic parameters in physical space. Also, the computational urden and local minimum are significantly less due to the sparseness and spatial relativity of the WSC. As an aid in prediction of significant coefficient, the idea of embedded zerotree wavelet (EZW) is imtroduced for the first time. Sequentially, the optimized parameter can be predicted using the zerotree and the process of inversion becomes better purpose.更多还原