Non-linear solution to seismic data conditioning using trained dictionaries

A tangible, non-transitory, machine-readable media, comprising instructions that when executed cause a processor to:

• receive a set of seismic data acquired via a seismic survey system, wherein the set of seismic data comprises at least one of undesired noise or undesired artifacts and further comprises a plurality of input volumes, wherein each respective input volume of the plurality of input volumes comprises positional data;
• select a first input volume of the plurality of input volumes having a first additional attribute as one of an offset, angle, or azimuth related to the first input volume;
• generate a pilot volume by selecting a range of input volumes of the plurality of input volumes based on attribute values of the plurality of input values and stacking input volumes of the range of input volumes with the first input volume;
• generate a trained dictionary based upon transformation of the pilot volume;
• transform the first input volume into transformed data;
• impose a sparse condition on the transformed data utilizing the trained dictionary to generate sparsified data;
• and
• inverse transform the sparsified data to generate an output data volume as a portion of a set of modified seismic data, wherein the output data volume has an improved signal-to-noise ratio relative to the set of seismic data.

A device, comprising:

• a processor that when in operation:
• receives a set of seismic data acquired via a seismic survey system, wherein the set of seismic data comprises at least one of undesired noise or undesired artifacts and further comprises a plurality of input volumes, wherein each respective input volume of the plurality of input volumes comprises positional data;
• selects a first input volume of the plurality of input volumes having a first additional attribute as one of an offset, angle, or azimuth related to the first input volume;
• generates a pilot volume by selecting a range of input volumes of the plurality of input volumes based on attribute values of the plurality of input values and stacking input volumes of the range of input volumes with the first input volume;
• generates a trained dictionary based upon transformation of the pilot volume;
• transforms the first input volume into transformed data;
• imposes a sparse condition on the transformed data utilizing the trained dictionary to generate sparsified data;
• and
• inverse transforms the sparsified data to generate an output data volume as a portion of a set of modified seismic data, wherein the output data volume has an improved signal-to-noise ratio relative to the set of seismic data.