Seismic time-frequency analysis with the deconvolutive W transform

Abstract

Time-frequency analysis is an effective method for interpreting seismic data and predicting hydrocarbon reservoirs. Among time-frequency analysis techniques, the W transform has proven effective in estimating the spectrum in low-frequency bands with higher resolution, and can better reveal reservoir characteristics in seismic data. It is a linear time-frequency analysis algorithm, defined as the convolution of a signal with a specific kernel function, where the analysis kernel acts as a low-pass filter and reduces the resolution of the W transform. We propose enhancing the resolution of the time-frequency spectrum by deconvolution. In this deconvolutive W transform, we estimate the point spread function as the Wigner-Ville distribution of the window function, for which we construct a sliding window, to accommodate the non-stationary window function, and determine the optimal window width and step size using Rényi entropy. The deconvolutive W transform retains the superior resolution of the standard W transform in low-frequency bands, while improving temporal resolution without compromising frequency resolution, and shows potential for thin-bed identification and low-frequency anomaly detection.