Seismic data Denoising and Reconstruction Based on Dictionary Learning Method
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Seismic exploration is a geophysical method to get the earth s subsurface layers information
newlineup to a certain depth, which helps to find the oil and minerals. In the seismic
newlineexploration method, a short-duration wave is sent from the earth s surface which is
newlineknown as a seismic wavelet. The seismic wave propagates inside the earth through
newlinesubsurface layers and gets reflected from the interface due to the impedance contrast
newlinebetween the two adjacent layers. These reflected waves are observed and recorded
newlineby an array of geophones. The recorded seismic data contains information about the
newlinesource wavelet, reflection coefficients and noises. The seismic data undergoes various
newlineprocessing steps like gain correction, deconvolution and stacking etc., to get the earth
newlinelayers information. After those processing steps, the seismic data retains reflection
newlinecoefficients with some level of noise. These noises severely effect the seismic imaging,
newlineinversion and interpretation.
newlineThere are two types of noise present in seismic data: coherent noise and incoherent
newlinenoise. Coherent noise occurs due to various factors such as surface waves, equipment
newlinemalfunctions, human activity, and multiple reflections of the seismic wavelet
newlinebetween layers. However, the incoherent noise is randomly distributed along the data
newlineand therefore it is difficult to estimate. The incoherent noise can be either Gaussian
newlinedistributed or non-Gaussian distribution. The noise which is uncorrelated from trace
newlineto trace and follows a Gaussian distribution is termed random noise. Where as the
newlineerratic noise is a term that is sometimes used to describe random incoherent noise
newlinethat does not follow a Gaussian or normal distribution. Attenuation of these random
newlineand erratic noises in the recorded seismic data is paramount importance in improving
newlinethe signal-to-noise ratio (SNR). However, attenuation methods that improve the SNR
newlinewhile preserving the useful signal are required for better interpretation of earth layers
newlineinformation.