Inversion-based 3D deblending of towed-streamer simultaneous source data using sparse TauP and wavelet transforms

We propose a 3D deblending method for towed streamer simultaneous source data based on an L1 inversion algorithm. It pursues the sparse representation of 3D coherent signals of the various sources that match the blended data in a combined transform domain of 2D TauP and 2D directional wavelet. In this method 2D sparse TauP is first applied to the data along consecutive channels for each shot, and then an L1 inversion algorithm based on 2D high angular resolution complex wavelet transform (HARCWT) is used to deblend the signals of different sources for each common P gather. Since the TauP transform along channels has good separation of the events in common shot gathers according to their slopes, and the coherent signals across shots have sparse representation in HARCWT domain, the method can achieve higher quality deblending results than 2D common channel deblending. The method was tested on numerically blended real field data, and the result was better than that of 2D deblending.