Wavelet transform domain filters: a spatially selective noise filtration technique

Area of Science:

• Signal Processing
• Image Analysis
• Medical Imaging

Background:

• Wavelet transforms offer multiresolution signal and image analysis.
• Edges are effectively detected in the wavelet transform domain.
• Noise reduction is crucial for accurate signal and image interpretation.

Purpose of the Study:

• To introduce a spatially selective noise filtration technique using wavelet transforms.
• To evaluate the technique's effectiveness in noise reduction and feature preservation.
• To compare the technique with existing methods like the Weiner filter.

Main Methods:

• Utilized direct spatial correlation of wavelet transforms at adjacent scales.
• Applied a high correlation threshold to identify significant features.
• Tested on simulated signals, phantom images, and real MR images.

Main Results:

• Achieved over 80% noise reduction in signals and images.
• Preserved at least 80% of edge gradient values.
• Demonstrated robustness with minimal artifacts, Gibbs' ringing, or resolution loss.

Conclusions:

• The developed wavelet-based noise filtration is highly effective and robust.
• It outperforms the Weiner filter in noise reduction and feature preservation.
• Potential applications include noise filtration, edge enhancement, image restoration, and artifact removal.