Multitrace Sparse-Spike Deconvolution Regularized by Spatial Derivatives

Sparse-spike deconvolution is a commonly used technology in enhancing the vertical resolution of post-stack seismic data. However, its inversion performance depends on the type and level of noise seriously. Besides, under classical trace-by-trace strategy the inverted reflectivity frequently suffer from poor continuity among traces. In this paper, we aim at alleviating above two issues, and try to obtain the laterally continuous reflectivity estimation from Gauss noise and non-Gauss noise background, respectively. To this end, a multitrace sparse-spike deconvolution method is tested. In this new approach we impose the Cauchy constraint along seismic traces to guarantee a full-band output. Simultaneously, the L1 norm constraint is placed on the spatial second-order gradients of the reflectivity to attenuate noise. A model and a partly real post-stack datum examples demonstrate that our improved method has an evidently better anti-noise ability compared to the single-trace sparse-spike deconvolution both in Gauss and non-Gauss noise cases. Moreover the new approach can readily explore more spatial continuities of geology structures.