|
<< Click to Display Table of Contents >> Navigation: Regularization > GTauP DeReverberation DeGhost |
The GTauP DeReverberation DeGhost module performs seismic data regularization and interpolation using a high-resolution 3D linear Radon (tau-p) transform. For each target bin, the module collects all input traces within a configurable spatial aperture, transforms the local gather into the tau-p domain, estimates and reconstructs the signal with sparse constraints, and then places the reconstructed trace back at the bin location. This process effectively fills missing offsets, regularizes irregular acquisition geometries, and can attenuate periodic multiples and ghost energy that appear as coherent but separated events in the tau-p domain.
The algorithm operates iteratively, applying a penalty factor to control the sparsity of the Radon-domain solution. A bandpass frequency filter is applied within the Radon transform to focus the estimation on the usable signal band and avoid fitting noise outside that range. The module supports distributed computation (remote execution), allowing large 3D datasets to be processed efficiently across multiple machines. Results are written to a new output seismic file in g-Platform's internal format.
This module is intended for prestack data regularization workflows, particularly where acquisition irregularities or cable feathering produce uneven offset or azimuth coverage. It is also useful as a preprocessing step before 5D interpolation, PSDM, or amplitude-versus-offset analysis that require well-sampled, uniform input gathers.
The input seismic dataset to be regularized. Connect the seismic file handle that contains the prestack data (for example, a common-midpoint or common-receiver sorted dataset). The module reads traces from this file to populate the local aperture gathers used by the Radon transform. The file must have valid trace headers so that source, receiver, and CMP coordinates can be used for spatial selection.
The sorted trace header index corresponding to the input seismic file. This index provides the geometry information (inline, crossline, source and receiver coordinates) that the module uses to locate traces within the spatial aperture around each output bin. The headers must be pre-sorted and matched to the seismic data handle connected above.
The path and file name for the regularized output seismic dataset. Specify a new file path with the .gsd extension (g-Platform internal format). If processing is interrupted and restarted, the module will check which bins are already written and skip them, allowing a safe resume of a partial computation.
The lateral search radius (in metres) used to select input traces contributing to each output bin. Only traces whose CMP location falls within this distance of the target bin are included in the Radon transform. Default: 5 m. Increase this value for sparse or coarsely sampled surveys to gather more contributing traces; reduce it for dense surveys to keep the aperture focused and avoid mixing arrivals from distant offsets. This parameter also defines the initial search radius used for geometry display in the interactive map view.
The aperture half-width (in metres) in the inline direction for selecting contributing traces during the Radon computation. Default: 5 m. This controls how far along the inline axis the module reaches when assembling the local gather. For a full 3D regularization effect, set the inline and crossline apertures to span at least one bin interval beyond the expected irregularity in the input geometry.
The aperture half-width (in metres) in the crossline direction. Default: 5 m. Works in conjunction with Aperture In line to define the rectangular spatial footprint from which input traces are drawn. The reference offset used internally for the Radon slant-stack normalisation is set automatically to the larger of the inline and crossline aperture values.
The time half-width (in seconds) of the sliding analysis window applied during the Radon transform. Default: 0.02 s (20 ms). A shorter window improves time resolution and adapts better to rapid changes in dip, but may reduce the accuracy of low-frequency components. A longer window increases statistical stability of the Radon decomposition at the cost of temporal resolution. This value is converted internally to the nearest number of samples based on the dataset sample interval.
The low-cut start frequency (in Hz) of the bandpass filter applied within the Radon domain. Default: 1 Hz. This is the frequency below which the filter begins to taper the signal toward zero. Set F1 lower than F2; together they define the low-frequency ramp of the bandpass. Frequencies below F1 are fully rejected from the Radon estimation.
The low-cut end frequency (in Hz) of the bandpass filter. Default: 3 Hz. Signal is passed at full amplitude at and above F2. The taper between F1 and F2 suppresses very low frequencies that could destabilise the inversion. For marine data where ocean swell noise is present below 3 Hz, keeping F2 at or above 3 Hz is recommended.
The high-cut start frequency (in Hz) of the bandpass filter. Default: 87 Hz. The filter begins to roll off above this frequency. Set F3 to match the upper edge of the usable signal band in your dataset. Frequencies above F3 and below F4 are progressively attenuated so that high-frequency noise does not contaminate the Radon inversion.
The high-cut end frequency (in Hz) of the bandpass filter. Default: 90 Hz. Frequencies at and above F4 are fully rejected. Ensure F4 does not exceed the Nyquist frequency of the dataset. The pair F3–F4 defines the high-frequency roll-off zone; a steeper roll-off (smaller gap between F3 and F4) gives a sharper high-cut but may introduce ringing.
The minimum ray parameter (slowness) value used to define the Radon transform space. Default: -0.025 s/m. The ray parameter p represents the slope of a linear moveout (dip) in the offset-time domain. Negative values correspond to energy dipping in the negative offset direction. Set P min to encompass the most negative moveout slope expected in your data. A range that is too narrow will clip steeply dipping events; a range that is unnecessarily wide will increase computation time.
The maximum ray parameter value. Default: 0.025 s/m. Together with P min, this value defines the full range of slownesses that the Radon transform will model. For typical reflection seismic surveys, a symmetric range of ±0.025 s/m covers moveout slopes corresponding to apparent velocities down to about 40 m/s over the aperture, which is sufficient for most primary reflections and moderate multiples. Adjust upward if very slow events (e.g. guided waves, surface-related multiples) must be included.
The ray parameter sampling interval used to discretise the Radon transform space. Default: 0.002 s/m. A finer Delta P (smaller value) produces a more densely sampled Radon domain, improving the ability to separate events with similar slowness, but significantly increases computation time. The same Delta P is applied in both the inline and crossline slowness directions. A good starting point is Delta P = (P max - P min) / 25.
The number of iterations used by the iterative high-resolution Radon inversion solver. Default: 10. More iterations allow the solver to progressively sharpen the Radon-domain model and better separate overlapping events, but increase computation time proportionally. For most regularization applications, values between 5 and 20 give a good balance of quality and speed. Use higher values (20–50) when events are very closely spaced in slowness or when artifact-free interpolation is critical for subsequent AVO or inversion workflows.
A damping or regularization factor applied to the iterative Radon inversion to control the sparsity of the solution. Default: 0.01 (1%). A larger penalty factor forces the Radon-domain model to be sparser (fewer active slowness components), which increases noise rejection at the expense of signal amplitude. A very small value (approaching zero) produces a least-squares solution with less noise attenuation. Start with the default and increase if coherent noise or multiple energy remains visible in the output; decrease if primary reflections appear over-attenuated.
A group of four parameters that restrict processing to a rectangular sub-volume of the survey, defined by inline and crossline number ranges. By default all four limits are set to -1, which means no limit is applied and the entire survey is processed. Use these parameters to test the module on a small representative section before committing to a full-volume run, or to restrict output to a specific area of interest.
The first inline number to include in the processing. Default: -1 (no lower limit). When set to -1, processing begins from the first inline present in the dataset. Set to a positive integer to skip earlier inlines and start processing from a specific inline number.
The last inline number to include in the processing. Default: -1 (no upper limit). When set to -1, processing continues to the last inline in the dataset. Set to a positive integer to stop processing at a specific inline number, which is useful when running tests or processing in sections.
The first crossline number to include in the processing. Default: -1 (no lower limit). Works together with Last crossline number to restrict the processed area in the crossline direction. Set to a positive integer to begin from a specific crossline.
The last crossline number to include in the processing. Default: -1 (no upper limit). When set to -1, all crosslines through the end of the survey are processed. Set to a specific crossline number to limit processing to a sub-area, for example when testing parameter settings or processing a pilot 3D corridor.
Refreshes the interactive acquisition geometry map displayed in the module's map view panel. Click Create maps after loading or changing the input data to update the display of all CMP bin locations. Once the map is populated, you can click on any bin to preview the Radon-based reconstruction result for that location before running the full processing job. The interactive display shows the input gather, the reconstructed (interpolated) gather, and the spatial aperture of contributing sources and receivers for the selected bin.