Seismic resolution enhancement using scale transform in the time-frequency domain

Abstract

Reducing uncertainty in geologic interpretation of petroleum reservoir containing thin layers requires increasing vertical resolution via appropriate advanced resolution enhancement methods. This problem was resolved here by introducing an alternative approach in resolution enhancement. Our method uses Gabor deconvolution (GD) combined with wavelet scaling. First, the seismic trace is transformed in the time-frequency domain using the Gabor transform. Subsequently, the Gabor magnitude spectrum of the seismic trace is smoothed to estimate the wavelet magnitude, which is then divided by the original value on the Gabor magnitude spectrum along the frequency axis to design a scale transformation filter. Finally, the filtered Gabor magnitude spectrum of the seismic trace is transformed back to the time domain using an inverse Gabor transformation. The result of using this technique is an increase of the dominant frequency, which produces a higher resolution of the seismic trace compared with using only the original GD. This method was applied to two synthetic and one field seismic data sets and compared with using only a GD. After applying the new approach, all three data sets indicate an extension in bandwidth (BW) and an enhancement in the resolution adequate for thin-layer seismic interpretation. When compared with the data sets using only a GD, the new approach produced comparable extension in the effective BW, while it pushing the dominant frequency to higher values. This allowed the imaging of many thin layers and geologic intervals in the field data example that could not be interpreted by the GD method.

Seismic Survey in Lesser Himalayan Thrust Belt, Western Nepal

“…A variety of VRE methods and strategies are available to solve the problem of thin-layer interpretation in seismic data [38]. Alaei et al [39] proposed a novel strategy of nonstationary scale transformation filter for VRE. The application of this method to three seismic data sets showed that it not only increases the bandwidth frequency of the data but it also increases the dominant frequency of bandwidth compared with the Gabor deconvolution method.…”

Fluid identification by nonlinear frequency‐dependent amplitude variation with offset inversion based on scattering theory

“…Although many linear approximation equations have been proposed, in order to improve the accuracy of reservoir fluid prediction, the high-order approximate scattering coefficient equation is derived using the medium decomposition theory. Through analysis, Aki and Richards (2002) gave the expression of scattering matrix in the case of plane wave incident as…”

“…Based on the exact Zoeppritz equations (Zoeppritz, 1919), AVO technology can predict reservoir parameters using the Bayesian estimation (Li et al, 2020;Wilson et al, 2009). Due to the complex structure of the exact Zoeppritz equations, it is difficult to apply them directly (Aki & Richards, 2002;Lan & Zong, 2022;Shuey, 1985). As a result, numerous researchers have attempted to simplify and approximate the Zoeppritz equation in diverse ways to tackle this challenge.…”

“…As a result, numerous researchers have attempted to simplify and approximate the Zoeppritz equation in diverse ways to tackle this challenge. One of the most commonly employed approaches is the Aki-Richards approximation (Aki & Richards, 2002), which is a linear approximation equation relating the velocities of the P-and S-waves and density. Based on this foundation, Fatti et al (1994), Rüger (1997), Shuey (1985), Zong et al (2013a) and others have developed linear approximations for Poisson's ratio, Young's modulus, wave impedance and fluid factor, respectively.…”

Depthwise separable convolution Unet for 3D seismic data interpolation

“…Seismic data reconstruction is important in seismic imaging. Many traditional intelligent methods have been proposed to solve various problems in seismic imaging, such as seismic resolution enhancement (Alaei et al, 2018;Soleimani, 2016;Soleimani, 2017;Soleimani et al, 2018;Mahdavi et al, 2021), complex geological structure identification (Soleimani, 2015;Farrokhnia et al, 2018;Kahoo et al, 2021;Khasraji et al, 2021;Hosseini-Fard et al, 2022;Khayer et al, 2022a;Khayer et al, 2022b), and noise attenuation (Gholtashi et al, 2015;Siahsar et al, 2017;Anvari et al, 2018;Anvari et al, 2020). In recent years, deep learning has achieved outstanding successes in a variety of domains, including computer vision (Ferdian et al, 2020;Manor and Geva, 2015) and medical image processing (Li et al, 2021;Tavoosi et al, 2021), with its powerful representing ability.…”