|
<< Click to Display Table of Contents >> Navigation: Filters > SVD De - Noise Filter |
This module is deprecated. Use the Cadzow De-Noise Filter or other noise attenuation modules for new processing workflows.
The SVD (Singular Value Decomposition) De-Noise Filter attenuates random and coherent noise from seismic gathers by applying a sliding 2D window SVD decomposition in the frequency domain. The filter retains only the most significant singular values, which correspond to coherent signal energy, and discards lower-ranked components that represent noise. The processing is performed trace-by-trace within a configurable spatial window, and the frequency range can be restricted to protect signal bandwidth.
The seismic data container connecting to the dataset to be filtered.
The input seismic gather to be de-noised. Works on any gather type (CMP, shot, receiver).
The half-width of the sliding trace window (in number of traces) used in the SVD computation. A larger window includes more traces in each local SVD, which can improve noise suppression for coherent noise but may reduce lateral resolution. Default: 40 traces.
The length of the time analysis window (in seconds) over which the SVD decomposition is performed. Longer windows capture more signal structure but may reduce vertical resolution. Default: 0.5 s.
The minimum frequency (Hz) of the band passed through the SVD filter. Frequencies below this value are left unchanged. Default: 1 Hz.
The maximum frequency (Hz) of the band processed by the SVD filter. Frequencies above this value are left unchanged. Default: 80 Hz.
The number of dominant singular values (eigenvalues) to retain in the reconstruction. Setting this to 1 keeps only the most coherent component; higher values preserve progressively more signal structure. Increasing this value reduces noise suppression but preserves more lateral variability in the data. Default: 1.