Noise attenuation

The Noise attenuation group contains modules for suppressing coherent and random noise in pre-stack and post-stack seismic data without muting. These modules apply spatial prediction, filtering, and wave-separation methods in the f-x, f-xy, tau-p, and wavelet domains to separate the desired signal from noise types such as ground roll, random ambient noise, spikes, and ringing. They are used throughout the processing sequence to improve the signal-to-noise ratio before velocity analysis, migration, and attribute analysis.

The following modules are available in this group:

2D Radial trace denoise — attenuates coherent noise (ground roll, direct wave) in 2D data by transforming each gather to the radial trace domain, where noise appears as low-frequency linear events, and filtering it there.

3D Radial trace denoise — extends radial trace noise attenuation to 3D gathers, suppressing azimuth-dependent coherent noise using radial trace filtering in 3D offset-time space.

3D volume radial trace denoise — applies radial trace noise attenuation to a full 3D post-stack or pre-stack volume, processing multiple inline-crossline positions simultaneously.

Adaptive ground roll attenuation — attenuates ground roll in land data using an adaptive filter that models the ground roll wavefield and subtracts it from each shot gather.

Despike — detects and removes large-amplitude spike noise from seismic traces using statistical outlier detection and sample interpolation.

Enhanced Butterfly — applies an enhanced butterfly (anti-aliasing) filter in the f-k domain to suppress aliased noise and migration artefacts in pre-stack or post-stack data.

FDNA — Frequency Domain Noise Attenuation: attenuates random noise by applying a spatially adaptive prediction filter in the frequency domain using f-x deconvolution principles.

FDNA plus — an enhanced version of FDNA with additional adaptive filtering options for higher-quality random noise attenuation on structurally complex data.

FX-Decon filter — attenuates random noise using f-x deconvolution, applying a Wiener prediction filter trace by trace in the frequency domain to enhance lateral coherence of reflectors.

FXY-Decon filter — extends f-x deconvolution to 3D data (f-xy), applying 2D spatial prediction in the frequency domain to suppress random noise while preserving coherent 3D reflectors.

LNA — Local Noise Attenuation: reduces spatially localised noise events using a local dip-guided filter that preserves signal dip while suppressing anomalous amplitude spikes.

LNA by local radon — performs local noise attenuation using a Radon-based dip decomposition within a sliding spatial window, providing higher-resolution noise suppression than standard LNA.

RNA (Random Noise Attenuation) — attenuates random noise using a statistical approach that models signal coherence across neighbouring traces and suppresses incoherent energy.

Trace mixing — mixes neighbouring traces using a weighted average to enhance lateral coherence and reduce trace-to-trace amplitude variations; useful for post-stack noise suppression and display enhancement.

Wave separation (HRWS) — High-Resolution Wave Separation: decomposes seismic gathers into upgoing and downgoing (or signal and noise) wavefields using a high-resolution dip-slowness filter.

Wave separation by gather (LPWD) — Low-frequency and P-wave Wavefield Decomposition: separates P-wave signal from low-frequency coherent noise and surface waves on a gather-by-gather basis using local plane-wave decomposition.