Suppressing random noise in seismic signals using wavelet thresholding based on improved chaotic fruit fly optimization

Suppressing random noise in seismic signals is an important issue in research on processing seismic data. Such data are difficult to interpret because seismic signals usually contain a large amount of random noise. While denoising can be used to reduce noise, most denoising methods require the prior estimation of the threshold of the signals to handle random noise, which makes it difficult to ensure optimal results. In this paper, we propose a wavelet threshold-based method of denoising that uses the improved chaotic fruit fly optimization algorithm. Our method of selects uses generalized cross-validation as the objective function for threshold selection. This objective function is optimized by introducing an adjustment coefficient to the chaotic fruit fly optimization algorithm, and the optimal wavelet threshold can then be obtained without any prior information. We conducted denoising tests by using synthetic seismic records and empirical seismic data acquired from the field. We added three types of noise, with different average signal-to-noise ratios, to synthetic seismograms containing noise with original intensities of - 5, - 1, and 4 dB, respectively. The results showed that after denoising, the signal-to-noise ratios of the three types of noise increased to 7.12, 10.04, and 14.26, while the mean-squared errors in the results of the proposed algorithm decreased to 0.006, 0.0031, and 0.0012, respectively.