Multiresolution generalized N dimension PCA for ultrasound image denoising

Area of Science:

• Medical imaging
• Image processing
• Signal processing

Background:

• Ultrasound images are inherently degraded by speckle noise, a form of random multiplicative noise.
• Speckle noise reduction is crucial for accurate medical diagnosis and interpretation of ultrasound data.

Purpose of the Study:

• To introduce a novel noise reduction technique for medical ultrasound images.
• To improve the visual quality and diagnostic utility of ultrasound images by reducing speckle noise.

Main Methods:

• A new method, multiresolution generalized N dimension PCA (MR-GND-PCA), is proposed.
• The method utilizes Gaussian pyramids and multiscale image stacks.
• Denoising is performed using generalized N dimension PCA (GND-PCA) at each level, with results combined via Laplacian pyramids.

Main Results:

• The MR-GND-PCA method was evaluated on synthetic and real ultrasound images.
• Performance was assessed using Mean Squared Error (MSE), Signal-to-Noise Ratio (SNR), and Peak Signal-to-Noise Ratio (PSNR).
• The method demonstrated effective noise reduction and structural preservation.

Conclusions:

• The proposed MR-GND-PCA method significantly reduces speckle noise in ultrasound images.
• It improves image quality while preserving essential structural details, crucial for clinical applications.
• The method shows robustness even with high levels of speckle noise.