Partial differential equation-based approach for empirical mode decomposition: application on image analysis.

Area of Science:

• Signal Processing
• Image Analysis
• Applied Mathematics

Background:

• Empirical Mode Decomposition (EMD) lacks a robust theoretical framework, hindering its evaluation.
• Existing 2D EMD extensions are often inefficient and time-consuming.
• Partial Differential Equations (PDEs) offer a promising alternative for signal processing tasks.

Purpose of the Study:

• To propose and describe an extension of a PDE-based sifting process for 2D Empirical Mode Decomposition.
• To address the limitations of the original Huang's EMD method in two-dimensional applications.
• To evaluate the effectiveness of the PDE-based approach for signal and image decomposition.

Main Methods:

• Utilized a nonlinear diffusion-based filtering process derived from PDEs for mean envelope estimation.
• Extended a previously developed 1D PDE-based sifting method to a 2D framework.
• Applied the new 2D PDE-based approach to both signal and image decomposition tasks.

Main Results:

• The proposed 2D PDE-based sifting process demonstrates significant improvements over existing 2D EMD methods.
• The approach effectively decomposes various types of signals and images.
• Successful application in image decomposition tasks was shown, highlighting its utility.

Conclusions:

• The PDE-based sifting process provides a theoretically sound and efficient alternative for 2D EMD.
• This method offers a valuable tool for signal and image processing applications like denoising, detrending, and texture analysis.
• The effectiveness of the 2D PDE-based approach warrants its further application in diverse data analysis scenarios.