Abrupt feature extraction via the combination of sparse representations

The complexity of channels in 3D seismic data always makes detailed interpretation challenging. Similarly, definition of sand bars and beaches is also complicated as they are only partially visible when seismic amplitude is examined. To improve the imaging of these features, current interpretation workflows use advanced color and opacity based co-rendering techniques to merge multiple attributes information. Other than to enhance these sedimentary features in 3D views with combinations of different attributes, this paper proposes to separate their reflection waveforms directly from 3D imaging data by exploiting the waveform diversity mechanism. Our separation model is set up upon the assumption that seismic data are composed of coherent events and abrupt features correspond to sedimentary features. According to their appearance in vertical sections, we model these two kinds of seismic features as linear structures and punctate structures respectively. Two appropriate waveform dictionaries are chosen, one of which is used for the representation of the coherent events and the other for the sedimentary features. The separation process is promoted by the sparsity of both waveform components in their corresponding representing dictionaries. The capacity of the proposed method is illustrated using modeling data and real 3D seismic data with complex depositional systems.