A Diffraction Separation Method using Variational Mode Decomposition

Summary Diffracted wavefields with superior illumination encode key geologic information about small-scale geologic discontinuities or inhomogeneities in the subsurface and thus possesses great potential for high-resolution imaging. However, the weak diffracted wavefield is easily masked by the dominant reflected data. Diffraction separation from specular reflected data still plays an important role and occupies a major position in diffraction implementation. To solve this problem, a new diffractionseparation method is proposed that uses variational mode decomposition (VMD) to suppress reflected data and separate diffracted wavefields in the common-offset or poststack domains. The VMD algorithm targeting reflected contributions decomposes seismic data into an ensemble of band-limited intrinsic mode functions representing linear and strong reflected data. The method based on the principle of energy sparsity can utilize the kinematic and dynamic differences between reflected and diffracted wavefields to effectively predict linear reflected data. Synthetic and field data examples using complex body geometries demonstrate the effectiveness and performance of the proposed method in enhancing diffracted wavefield and attenuating reflected data as well as increasing the signal-to-noise ratio, which helps to clearly image small-scale subwavelength geologic structures.