Orientation-Independent Empirical Mode Decomposition for Images Based on Unconstrained Optimization

Area of Science:

• Signal Processing
• Image Analysis
• Data Science

Background:

• Empirical Mode Decomposition (EMD) is a data-driven technique for analyzing complex signals.
• Existing 2D EMD methods can be computationally intensive and orientation-dependent.
• Decomposition of Amplitude Modulated-Frequency Modulated (AM-FM) images presents unique challenges.

Purpose of the Study:

• To introduce a novel, computationally efficient, and orientation-independent 2D extension of Empirical Mode Decomposition (EMD).
• To improve the decomposition performance for Amplitude Modulated-Frequency Modulated (AM-FM) images.
• To demonstrate the method's applicability in biomedical data analysis.

Main Methods:

• A new 2D Empirical Mode Decomposition (EMD) approach based on unconstrained optimization is proposed.
• The method is implemented for fast and simple signal separation into oscillatory components.
• Performance is evaluated against state-of-the-art techniques.

Main Results:

• The proposed 2D EMD method demonstrates superior computational efficiency and orientation independence.
• It achieves better performance in decomposing Amplitude Modulated-Frequency Modulated (AM-FM) images compared to existing methods.
• Successful application on artificial AM-FM images and a biomedical dataset is shown.

Conclusions:

• The novel 2D EMD offers a significant advancement in signal processing for image analysis.
• The method is robust, efficient, and adaptable for various applications, including biomedical imaging.
• The framework supports potential extensions to n-dimensional data (n > 2).