The Application Research Of Matching Pursuit In Seismic Signal Denoising

To improve the resolution of seismic data is a main task in the seismic data processing, and its prerequisite condition is to improve the signal-to-noise ratio. This paper is the preliminary application in denoising of seismic signal with matching pursuit algorithm.Firstly the paper introduces the characteristic of the noises in seismic data, and then it introduces the denoising methods in coherent noises and non-coherent noises, and analyses the advantages and disadvantages of these methods.Then, the basic theory of sparse decomposition is introduced, and dose detailed explanation to the theory and implementation of the matching pursuit algorithm. Applying to the application of MP in seismic signals'decomposition, through some analyses of seismic signal's characteristics in the paper, the non-zero phase's Ricker wavelet is selected to establish the over-complete dictionary.The method of Matching Pursuit algorithm based on frequency filtering is provided in order to wipe out surface wave interferences through the study of this noise's characteristics. This method is based on the frequency property of surface wave interferences, a threshold frequency is selected by experience value, and then the time-frequency atoms whose frequency is less than the threshold will be removed so as to reach the purpose of wiping out the surface wave interferences. According to the characteristic of MP algorithm, the speed of this algorithm is decreased exponentially with the increase of the signal's length. Therefore, in order to further improve the algorithm's speed, two kind of adding window methods is discussed, a self-adapt adding window method is provided through reducing the signal's length in every step to improve the whole speed of the algorithm.Finally, this paper proposes a MP method based on dip scanning to remove linear interference. This method makes full use of the linear noises'characteristics. Firstly, it through dip scanning and small window techniques to cut out the seismic data into a small area to denoise. Then, through the correlation analysis to determine whether the area's data contain linear interference. If the average correlation coefficient of this data is larger than threshold value, then there is linear noises in the data which need to be denoised. Base on the frequency and linear property of linear interference, frequency method and translation method is provided in the paper. One is based on the frequency parameter of time-frequency atoms to judge whether it meets the characteristics of linear interference, and the other one is based on the translation parameter. Linear noises can be wiped out through these two methods. Since the algorithm is also applied to the small window technology, the detailed analyses have been done about how to select the time window. There are lots of simulation experiments on every algorithm, and the results are analyzed in detail. These simulation experiments have verified the efficiency of these methods.