Hybrid Demultiple Method in the Radon Domain

In the shallow water environment predictive deconvolution (PD) is an effective method for predicting water-layer multiples and reverberations (WLMRs). After that the adaptive multiple subtraction in the time-offset domain (TOAMS) is usually exploited. Then the residual multiples, such as ineffectively predicted long period multiples, can be further suppressed by the parabolic Radon transform method (PRT). However, multiples that are partially subtracted by the TOAMS and have residual moveout close to that of primaries are still left. In this paper we propose the hybrid demultiple method combining the PRT and AMS in the Radon domain. After predicting WLMRs with PD, we first divide the Radon image of original data into three areas. We mute the Radon area that primaries map into, and keep the Radon area that ineffectively predicted multiples map into. For other area we use the 2D matching filter to estimate multiples. At last the estimated multiples in the Radon domain are transformed to the time-offset domain and subtracted from the original data. Compared with the cascaded method with the TOAMS followed by the PRT, it can remove the multiples that are partially subtracted by the TOAMS and have residual moveout close to that of primaries.