An empirical mode decomposition based noise cancelation method for potential field data along with a new stopping criterion

Potential field data is generally contaminated by random noise. The high-frequency noise contained in the data brings unfavorable influences to subsequent data processing. Therefore, suppressing the adverse effects of noise has always been a crucial step which is desirable prior to applying other transformations. Over the past decades, numerous mathematical approaches have been proposed for noise cancelation of potential field data. In the work discussed in this paper, the application of the empirical mode decomposition for denoising of potential field data is briefly described, and a new stopping criterion for this filtering method is introduced. Using the proposed method, the empirical mode decomposition is firstly performed on the original potential field data to get numerous intrinsic mode functions corresponding to components with different frequencies. Each intrinsic mode function is subtracted from the original data to get different residual datasets. The correlation coefficients associated with the original data and various residual datasets are calculated and plotted. The inflection point of the correlation coefficient curve is adopted as the last intrinsic mode function to be selected. The new stopping criterion offers a quantitative way to determine which intrinsic mode functions should be removed during filtering and can be easily implemented within the algorithm. Tests on synthetic noisy gravity data demonstrate that the empirical mode decomposition based noise cancelation method along with this new stopping criterion yield acceptable filtering results for potential field data. The newly developed method is also investigated on real gravity data collected over a magnetite zone in Jilin Province, China.