Noise Suppression
SEARCH
FK Filtering
FK dip filtering is performed by dividing the FK spectrum of the data into pass and reject zones. Either a single pass-range of dips can be specified, or a ‘pie-slice’ division into pass or reject zones can be specified. In all cases, low frequency protection is under user control. Input data can be any multi-trace group, such as shot records, receiver gathers, common offsets etc. The FK transform is computed assuming regular trace spacing.
Tau-p Filtering
Tau-p filtering/muting can be used to attenuate noise by dividing the Tau-p transform into pass and reject zones. This is similar to an FK filter, in that the p-traces correspond to dip as in the FK filter. However, the reject zone can now be made time dependent. The reject zone may be any shape. A linear taper between the pass and reject zones is applied. It is also possible to scale p traces. Band-pass filters can also be applied in Tau-p to attenuate different frequencies at different dips and times. The Tau-p transform can be computed for either regular or irregular input trace spacing.
FXY Deconvolution
FXY deconvolution is the 3D extension of FX deconvolution. A 1-pass FXY filter is designed and applied to 3D data. The user controls the spatial and temporal windows used for filter design and the amount of signal add back. The module can be applied to any 3D dataset such as stacks, common offsets or cross-spreads.
Swell Noise Attenuation
Swell noise is caused by wave action in marginal weather conditions and is typically characterised by high amplitude bursts of incoherent noise over a range of channels in marine seismic data. Some other sources of noise, such as fish bites, can have similar characteristics. Swell noise attenuation methods typically use the spatial and spectral characteristics of the noise to identify and attenuate it while preserving data amplitudes of the underlying geology. Successful techniques have included XWAVE and FXEDIT based methods.
De-spiking
De-spiking is used for the suppression of impulsive noise by the detection and suppression of high amplitude impulsive noise. Within a time and space window, a rolling median amplitude is computed. If any component is greater than the median by a user specified threshold, it is deleted and either zeroed or scaled back to the median value. The module can be applied within any multi-trace ensemble.
Similar approaches can be used to compare amplitudes above and below a low frequency threshold so as to better target and attenuate predominately low frequency noise such as swell noise and ground roll.
Spectral Editing: FXEDIT
FXEDIT is used for the suppression of impulsive noise by the detection and repair of anomalously high amplitude windows of a trace by FX projection filtering. Within a time and space window and frequency-by-frequency a rolling median amplitude is computed. If any frequency component is greater than the median by a user specified threshold, it is deleted and reconstructed using an FX prediction filter. The module can be applied within any multi-trace ensemble.
Shots before FXEDIT
Shots after FXEDIT
High-Resolution Linear Noise Attenuation: XRLIN
High-Resolution Linear Noise Attenuation: XRLIN
XRLIN performs noise removal and trace regularisation using a constrained, high-resolution, linear Radon transform. The algorithm gives a better-focused representation of the data in the Radon domain compared to conventional least-squares transform or to direct summation (as in standard -p applications). The program is better able to preserve primary amplitudes as a function of offset whilst simultaneously giving a more complete noise removal than those approaches. It is also resistant to spatial aliasing and can therefore reduce the need for trace interpolation before noise removal. The algorithm honours the true offset of the data, so that linear noise trains can be accurately attenuated even in irregularly sampled land or sea floor acquisition.
This application can also be used to generate output traces at different locations from the input traces, by specifying a different set of offsets for the inverse transform. If the input traces are irregularly sampled, for example, then the output, with or without noise removal, can be generated at a regularised set of trace locations.
The following examples show the effects of applying XRLIN to the shots. Example shots have been displayed on the following pages.
Shots – Pre XRLIN
Shots – Post XRLIN