An improved bidimensional empirical mode decomposition: A mean approach for fast decomposition

• A novel mean approach is proposed to accelerate the bidimensional empirical mode decomposition.
• The proposed method utilizes a singular-value-decomposition-based convolution algorithm to reduce computation time.
• The distance computation is reduced by Delaunay triangulation.
• Experimental results show the proposed method is faster than a state-of-the-art method.
• Experimental results show the proposed method keeps acceptable quality.

In this paper, a mean approach is proposed to accelerate bidimensional empirical mode decomposition (BEMD). In the envelope generation process, the proposed method uses a modified mean filter to approximate the interpolated envelope of the conventional BEMD, and utilizes a convolution algorithm based on singular value decomposition (SVD) to further reduce the computation time. Order statistics filter width determination, originally used in fast and adaptive bidimensional empirical mode decomposition (FABEMD), is applied to adaptively formulate an envelope. Considering the computation efficiency, the proposed method improves the algorithm for calculating distances among extrema by using Delaunay triangulation (DT). The experimental results show that the mean approach can produce intrinsic mode functions faster than FABEMD, while retaining acceptable quality.