Deblending using dictionary learning with virtual shots

A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:

• generating, from blended shot data of a seismic survey of an area that includes signal data and noise data, a shot domain for actual sources, a receiver domain for virtual sources, and a receiver domain for actual sources, wherein the shot domain includes shots provided by an array of vibration-emitting equipment, each piece of vibration-emitting equipment being a single source in the seismic survey, wherein the receiver domain for the actual sources includes shot gathers received at receiver locations, and wherein each shot gather includes signal atom data gathered during a small time window using multiple neighboring traces in a seismic gather;
• generating, during a first dictionary learning process using the shot domain for the actual sources, a dictionary of signal atoms, wherein each signal atom includes a small patch of seismic signal data gathered during the small time window using multiple neighboring traces in the seismic gather, wherein the first dictionary learning process includes applying a time shift to one or more of the shot domain for the actual sources, wherein the first dictionary learning process identifies a set of similar signal atoms with which their respective inputs are reconstructed, and wherein a gather is one of a common source, a common receiver, a common offset, or a common midpoint gather, and wherein the small patches define time and space windows of 100 milliseconds of trace data across 20 consecutive traces;
• generating, during a second dictionary learning process using the receiver domain for the virtual sources, a dictionary of noise atoms, wherein the second dictionary learning process includes applying a time shift to the receiver domain for the virtual sources, and wherein each noise atom includes a small patch of seismic noise data gathered during a small time window using multiple neighboring traces in a seismic gather;
• generating a combined signal-and-noise dictionary that contains the signal atoms from the dictionary of signal atoms and the noise atoms from the dictionary of noise atoms;
• creating a sparse reconstruction of receiver domain data from the combined signal-and-noise dictionary;
• and
• splitting the sparse reconstruction into deblended data and blending noise data based on atom usage to create deblended shot domain gathers for the actual sources.

A computer-implemented system, comprising:

• one or more processors;
• and
• a non-transitory computer-readable storage medium coupled to the one or more processors and storing programming instructions for execution by the one or more processors, the programming instructions instructing the one or more processors to perform operations comprising:
• generating, from blended shot data of a seismic survey of an area that includes signal data and noise data, a shot domain for actual sources, a receiver domain for virtual sources, and a receiver domain for actual sources, wherein the shot domain includes shots provided by an array of vibration-emitting equipment, each piece of vibration-emitting equipment being a single source in the seismic survey, wherein the receiver domain for the actual sources includes shot gathers received at receiver locations, and wherein each shot gather includes signal atom data gathered during a small time window using multiple neighboring traces in a seismic gather;
• generating, during a first dictionary learning process using the shot domain for the actual sources, a dictionary of signal atoms, wherein each signal atom includes the small patch of seismic signal data gathered during a small time window using multiple neighboring traces in the seismic gather, wherein the first dictionary learning process includes applying a time shift to one or more of the shot domain for the actual sources, wherein the first dictionary learning process identifies a set of similar signal atoms with which their respective inputs are reconstructed, and wherein a gather is one of a common source, a common receiver, a common offset, or a common midpoint gather, and wherein the small patches define time and space windows of 100 milliseconds of trace data across 20 consecutive traces;
• generating, during a second dictionary learning process using the receiver domain for the virtual sources, a dictionary of noise atoms, wherein the second dictionary learning process includes applying a time shift to the receiver domain for the virtual sources, and wherein each noise atom includes a small patch of seismic noise data gathered during a small time window using multiple neighboring traces in a seismic gather;
• generating a combined signal-and-noise dictionary that contains the signal atoms from the dictionary of signal atoms and the noise atoms from the dictionary of noise atoms;
• creating a sparse reconstruction of receiver domain data from the combined signal-and-noise dictionary;
• and
• splitting the sparse reconstruction into deblended data and blending noise data based on atom usage to create deblended shot domain gathers for the actual sources.