Rayleigh-Rice Mixture Parameter Estimation via EM Algorithm for Change Detection in Multispectral Images.

Area of Science:

• Image Analysis
• Remote Sensing
• Medical Image Processing
• Change Detection

Background:

• Estimating Rayleigh-Rice mixture density parameters is vital for image analysis tasks like remote sensing and medical imaging.
• Change detection (CD) in multispectral images often uses change vector analysis, where the difference image magnitude follows a Rayleigh-Rice mixture density.
• Approximating this complex model with Gaussian mixtures can reduce CD accuracy.

Purpose of the Study:

• To develop a novel and effective technique for estimating Rayleigh-Rice mixture density parameters.
• To apply this technique to change detection in multitemporal and multispectral images.
• To demonstrate the superiority of the theoretically derived Rayleigh-Rice model over empirical models in CD.

Main Methods:

• A novel parameter estimation technique for Rayleigh-Rice mixture density based on a specific expectation-maximization (EM) algorithm.
• Iterative parameter updates without reliance on specific optimization routines.
• Validation through numerical experiments on synthetic and real remote sensing data.

Main Results:

• The proposed EM-based method effectively estimates Rayleigh-Rice mixture density parameters.
• The theoretically derived Rayleigh-Rice model significantly outperforms empirical models in change detection.
• Experiments on real remote sensing data show substantially higher CD accuracies compared to state-of-the-art methods.

Conclusions:

• The novel EM algorithm provides a robust and theoretically sound method for Rayleigh-Rice mixture density parameter estimation.
• This technique enhances change detection accuracy in multispectral and multitemporal remote sensing.
• The method is general and applicable to various image processing problems involving Rayleigh-Rice mixture densities.