Seismic attenuation compensation by Bayesian inversion

•Improve the compensation stability by Bayesian inversion.•Choose proper frequency band to avoid high frequency noises.•Attenuation compensation without knowing the accuracy Q value

As an effective method to improve seismic data resolution, attenuation compensation has been paid great attention. The popular method inverse Q-filter performs effectively in phase correction. But its energy compensation part is at an extraordinary discount because of its instability. By contrast, the compensation method based on inversion has a great advantage in algorithm stability. In this paper, the inversion process is combined with Bayesian principle so that the prior information we learned about the actual model can be used sufficiently. Here, we have an assumption that it is more reasonable to describe the reflectivities with sparse distribution. This information, in general, can be transferred to a sparse constraint of the object function. And compared with Tikhonov regularization method, it is proved to perform better in seismic resolution improvement. Meanwhile, it is insensitive to the error of Q value. Example of real data shows the validity of the method.