Method of removing incoherent noise
AssigneePGS Geophysical AS
InventorRoald Gunnar VAN BORSELEN
Life of this patent
Estimating the term…
Abstract
Noise is removed from a data set by performing an integral transform operation that converts instances of noise into identifiable artefacts in the transformed data set. A model of the artefacts is constructed by creating a full-domain or partial-domain noise model and performing the same integral transform operation on the noise model. The resulting transformation of the noise model is adaptively subtracted from the transformed data set to remove the noise. The adaptive subtraction may employ a least-square error filter.
Claims20
A method of removing noise from a data set, comprising:
- using a computer to perform a transformation on the data set, the transformation comprising:
- performing an operation on the data set that transforms incoherent energy features into artefacts having an identifiable structure to form a patterned data set, the operation comprising an integral transform;
- using the data set to construct a model data set containing incoherent noise from the data set that, when subjected to the operation, produces artefacts similar to the artefacts in the patterned data set;
- performing the operation on the model data set to form a noise model;
- constructing an energy minimizing filter from the patterned data set and the noise model;
- applying the energy minimizing filter to the noise model to form a filtered noise model;
- and
- subtracting the filtered noise model from the patterned data set to form a transformed data set.
A method of removing noise from a data set, comprising:
- performing a wavefield decomposition on the data set to form a wavefield patterned data set;
- constructing a model data set containing an incoherent noise sample;
- performing the wavefield decomposition on the model data set to form a noise model;
- constructing an energy minimizing filter from the wavefield patterned data set and the noise model;
- applying the energy minimizing filter to the noise model to form a filtered noise model;
- and
- subtracting the filtered noise model from the wavefield patterned data set to form a de-noised data set;
- and
- performing an inversion of the wavefield decomposition on the de-noised data set.
A method of removing noise from a data set, comprising:
- using a computer to perform a transformation on the data set, the transformation comprising:
- extracting coherent energy signals from the data set to form a coherent model;
- subtracting the coherent model from the data set to form an incoherent model;
- performing a wavefield decomposition on the data set to form a wavefield patterned data set;
- performing the wavefield decomposition on the incoherent model to form a noise model;
- constructing an energy minimizing filter from the wavefield patterned data set and the noise model;
- applying the energy minimizing filter to the noise model to form a filtered noise model;
- and
- subtracting the filtered noise model from the wavefield patterned data set to form a de-noised data set;
- and
- performing an inversion of the wavefield decomposition on the de-noised data set.
Dependencies are inferred from the claim text.