LAUSR

Abstract Traditionally, a denoising algorithm is usually applied to enhance useful reflection seismic signals, or to increase the quality of the waveforms recorded in the seismic data. The strong noise left in the filtered data after denoising is a long standing drawback of many state-of-the-art denoising algorithms. In addition to removing noise, we are proposing a new way to detect the waveforms, which can also enhance the useful signals. We propose a waveform classification algorithm based on a clustering method, e.g., to classify a data point in the data into signal or noise. The waveform classification algorithm can be applied directly to the noisy data to solely output the picked signals and thus can be a complete substitute to a traditional denoising algorithm. A better way for “denoising” is to pick the signals from a traditional denoising algorithm with the labels classified from the noisy data. At the same time, we also propose a novel automatic denoising method based on the empirical wavelet transform (EWT). We use the EWT method as a template to show how the proposed waveform classification method can be combined with a traditional denoising algorithm to output a superb denoised signal.