The sparse reflection coefficients-based seismic multichannel convolution model

This paper proposes a new multichannel convolution model for stacked seismic data. This model supposes that the layers are horizontal and the reflection coefficients are sparse. The properties of the layer change weakly and randomly, which means that the reflection coefficients from same interface change randomly with weak amplitude. Based on this model and lateral invariant seismic wavelet, we introduce the time-vary wavelet convolution model of multichannel seismic signal. Therefore the instantaneous linear mix model of independent component analysis (ICA) is satisfied. The time-vary wavelet can be estimated by solving proposed model. The synthetic and field data examples demonstrate the rationality of proposed model and the validity of the seismic wavelet estimated method. The wavelet estimated matched very well with the wavelet that was used in synthesizing data. Using the wavelet estimated by proposed method to deconvolve the real field data, the time-resolution was enhanced evidently, and the high signal-to-noise rate (SNR) was also kept.