A Theory for Multiresolution Signal Decomposition: The Wavelet Representation
IEEE Transactions on Pattern Analysis and Machine Intelligence
Ten lectures on wavelets
Orthonormal bases of compactly supported wavelets II: variations on a theme
SIAM Journal on Mathematical Analysis
An introduction to wavelets
Rician noise attenuation in the wavelet packet transformed domain for brain MRI
Integrated Computer-Aided Engineering
| Hi-index | 0.00 |
Seismic data is often contaminated with high-energy, spatially aliased noise, which has proven impractical to attenuate using Fourier techniques. This problem is compounded when the data volumes become large and the noise characteristics variable. Wavelet filtering has proven capable of attacking several types of localized noise simultaneously regardless of their frequencies. A stationary wavelet transform is used to decompose seismic trace data into its wavelet components; these are localized in both time and frequency. A threshold is applied to these coefficients to attenuate high amplitude noise, followed by an inverse transform to reconstruct the seismic trace. The wavelet-transform coefficients of a seismic trace describe the temporal and frequency (or spatial and wavenumber) distribution of the energy in the trace. The stationary wavelet transform minimizes the phase-shift errors induced by thresholding that occur when the conventional discrete wavelet transform is used. A land 3D seismic acquisition example is cited where both dynamite and vibroseis sources were employed. In these data, high-amplitude noise events are present, and there are vastly different energy levels between the two source types, in which the noise amplitudes vary by up to six orders of magnitude. A data-adaptive threshold determination process in wavelet filtering significantly increases the ratio of signal to noise.