LIFT: a new and practical approach to noise and multiple attenuation

Abstract

A New Approach Traditional noise attenuation methods attempt to separate signal from noise by transforming data into a domain where the signal or noise is modeled mathematically, and signal and noise can be separated. Many methods simply stop at separating noise and signal. That is, the signal model itself is the output of the noise attenuation program. Some compensate for the 'too clean' synthetic look of the output by adding back a percentage of the original input data. LIFT (Choo and Sudhakar, 2003), a new amplitude-friendly technique to attenuate noise and multiples, takes a new approach by adding back an estimate of the signal lost during the modeling, rather than simply outputting the signal model or the signal model with a percentage of the original data added back. This is a fundamental shift in noise attenuation strategy. This approach greatly improves signal preservation. It is very flexible in that it can incorporate a variety of application domains, filtering tools, and ways of modeling data, as appropriate to address particular noise problems. As well, it is a practical and robust amplitude-preserving way to precondition data for pre-stack migration, to prevent migration artifacts and costly reruns. Noise Attenuation Current Practices Many and varied strategies have been used to model the signal or noise, all making various assumptions. For instance, the K-L transform assumes that the signal is flat in time with only amplitude variations. The Radon transform is more flexible in that it allows signal to follow varied trajectories, but a few fixed types of trajectories (hyperbolic, parabolic, and straight lines for example) cannot fully cover all signal behavior. The F-X prediction filter is a popular tool to attenuate random noise. It makes a general signal model assumption: that signal is similar from one trace to the next and spatially predictable by convolution filters-a less restrictive signal model than that of some other noise attenuation algorithms.

Recovery of temperature, salinity, and potential density from ocean reflectivity

“…LIFT filtering . We have applied the LIFT approach [ Choo et al ., ] to increase the signal‐to‐noise ratio and to filter out the direct wave. LIFT is an amplitude‐preserving noise attenuation algorithm composed of a sequence of models applied to data divided in frequency windows and sometimes also to source‐receiver offset domains.…”

Seismic reflection imaging of the Juan de Fuca plate from ridge to trench: New constraints on the distribution of faulting and evolution of the crust prior to subduction

“…The MCS data were processed using the following sequence: a pseudo 3‐D geometry was defined to minimize the effects of streamer feathering, with exact source‐receiver offsets preserved and data binned into 6.25 m long common midpoint (CMP) gathers along track; band‐pass filtering (3‐7‐220‐250 Hz) was applied to remove cable noise, followed by trace editing, spherical divergence correction, and resampling to 4 ms. Due to the electrical issues on board Langseth at the time of our survey, some channels are noisy with amplitude spikes. To enhance the signal‐to‐noise ratio and to avoid the smearing in later migration, despiking with the LIFT method [ Choo et al ., ] was applied (Figure S1). In our application of the LIFT method, the data were first divided into three frequency bands (low‐frequency band: 0–14 Hz, midfrequency band: 14–56 Hz, and high‐frequency band: 56–125 Hz).…”

Distribution of Crustal Melt Bodies at the Hot Spot‐Influenced Section of the Galápagos Spreading Centre From Seismic Reflection Images

“…(1) CMP geometry definition for all seismic data traces; (2) band-pass filtering (1-6-100-125 Hz); (3) noise attenuation using the LIFT method, a processing technique that suppresses noise while reconstructing signal to its original form preserving amplitude integrity Choo et al, 2004;Han et al, 2016; Figure S3B); (4) offset-dependent spherical divergence correction to compensate for geometrical spreading; (5) surface-consistent amplitude balancing to normalize abnormally high/low shot and channel amplitudes; and (6) seafloor primary multiple mute to reduce migration noise. Sinton et al, 2003) near the Wolf-Darwin and Central volcanic lineaments (Mittelstaedt et al, 2012).…”