Application of undecimated discrete wavelet transforms (UDWT) for seismic refraction velocity analysis improvement

Abstract

In this study, the wavelet transform is applied to the seismic refraction data to provide an accurate first break picking result. The goal is not to apply the threshold to filter the data for the random noise suppression. We tried to introduce the wavelet domain as a replacement for the time domain. Thus, changing the domain is so hazardous; because the kernel effects of each transform could change the first break, which means that an erroneous refraction interpretation but note that DWT approaches have divided into many types based on frames, filter Banks and dimensions. In this way, this study has found one of the suitable domains of DWT that could provide a reliable result. Another aspect of this study that should be mentioned is that there is no reconstruction performed after decomposition. Our finding shows that to provide a better decomposition for the first break, picking the wavelet type must be short enough to avoid the effect of the kernel selection. We are, therefore, test db2 as a mother wavelet. After the decomposition, the data will be provided to a number of scales. Each scale has its frequency content, which means that random noise concentration will be at particular scales. These scales that contain more random noise will be eliminated and the best scale with the lowest possible amount of random noise will be chosen by the statistical approaches. After this selection, the first break selection will proceed to provide accurate interpretations. Subsequently, the velocity of each layer will be estimated using these first arrival times. Another achievement in this study is to find a true scale in DWT domain to avoid redundancy aspects. To select the true scale, a statistical approaches proved that we can find best scale for analysis.