Technical Paper: Seismic Full-Waveform Inversion Using Truncated Wavelet Representations

Society: SEG
Paper Number: 2012-43
Presentation Date: 2012


We present a model compression scheme for solving acoustic full-waveform inversion problems using the Gauss-Newton minimization method. In this scheme, we represent P-wave velocity and mass-density distributions using wavelet basis functions. In order to reduce the number of unknown parameters in the inversion, we invert only dominant wavelet coefficients. Since the number of dominant coefficients is smaller than the number of unknown parameters in the spatial-domain, this compression scheme reduces the size of the Jacobian matrix. Hence, we reduce the memory storage of the Jacobian matrix as well as the computational time for calculating the Gauss-Newton update step. We use the Marmousi model to show that the model compression scheme can reduce the computational time and memory storage of the Gauss-Newton method without sacrificing the quality of its inversion results.

Request More Information