Post-processing (3D)#
Recommended post-processing steps applied to sparse 3D cube after POCS interpolation.
Description#
Inline/crossline upsampling#
One optional post-processing step is the upsampling of the inline or crossline bin size to match respective smaller bin size (Figure 1). The default approch is a simple linear interpolation, which is only sufficient for small bin size differences. Available interpolation methods are listed in the documentation for the xarray.Dataset.interp_like function.
Additionally, one can apply a filter to reduce the spatial aliasing if present using the --spatial-dealiasing switch.
Upsampling is triggered using the following command line parameter:
--upsample {linear (default) | nearest | slinear | cubic | polynomial}
--spatial-dealiasing # Whether to apply filter in kx-ky domain to remove spatial aliasing.
Acquisition footprint removal#
Acquisition footprint describes linear spatial grid patterns seen on 3D time slices (Figure 2). Commonly caused by the aquisition pattern and direction for multichannel seismic data, these patterns might be present in high-resolution SBP data as well.
Individual time/frequency slices are filtered by damping zero-centered amplitudes in the \(k_x-k_y\) domain after 2D FFT application (Figure 2). Our implementation allows the footprint removal over:
-
(time) slice
-
profile (automatic iline/xline detection)
-
profile-iline (over inlines)
-
profile-xline (over crosslines)
The removal of the acquisition footprint is triggered using the following command line parameter:
Smoothing of time/frequency slices#
The data can be filtered in the frequency domain to remove white noise (Figure 3). Two different filters are supported:
- gaussian:
--smooth gaussian --smooth-sigma {default: 1} - median:
--smooth median --smooth-size {default: 3}
Additionally, the smoothed slices can be rescaled to a specific percentile range of the input data (default: [0.01, 99.99]) using --rescale {0.01 99.99}.
Automatic Gain Control (AGC)#
One of the most common gain functions is the Automatic Gain Control (AGC) using time-variant scaling to increase weak amplitudes. AGC is applied to the data on a trace-by-trace basis using a sliding time window (Figure 4). The scale factor calculated for each sliding window can refer to the rms (default), mean, or median amplitude.
Warning
Be aware that the window length is in seconds!
Usage#
This script is designed to be used from the terminal (i.e. command line).
Command line interface#
The script needs a single netCDF (3D) as input:
There are two options to run the script. We recommend using the CLI entry point like:
Alternatively, the script can be executed using the (more verbose) command:
Optionally, the following parameters can be specified:
--help,-h: Show help.--path_out: Output path of pre-processed 3D cube.--upsample: Upsamplexarray.Datasetto equal bin size along ilines and xlines.linear,nearest,slinear,cubic, orpolynomial
--spatial-dealiasing: Whether to apply filter in kx-ky domain to remove spatial aliasing.--remove-footprint: Remove acquisition footprint (default:slice).--direction: Direction of acquisition footprint removal filter (default:both).--footprint-sigma: Standard deviation for smoothing Gaussian filter (default:7) to remove footprint.--buffer-center: Percentual buffer (0-1) around center in \(k_x-k_y\) domain (default:0.20).--buffer-filter: Footprint filter buffer size (in grid cells, default:3).--smooth {gaussian | median}: Smooth slices (frequency or time domain).--smooth-sigma: Standard deviation for Gaussian kernel (default:1).--smooth-sigma: Shape of Median kernel (identical for iline and xline) in grid cells (default:3).--rescale: Rescale smoothed slices to given percentile range (without arguments: [0.01, 99.99]).--verbose {LEVEL}: Level of output verbosity (default:0).