Compressive sensing based on L₁ and Hessian regularizations for MRI denoising.

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Science

Background:

• Compressive sensing (CS) is a technique used for signal acquisition and reconstruction.
• Traditional CS methods can lead to over-smoothing and loss of fine details during noise reduction.
• Magnetic Resonance Imaging (MRI) is susceptible to noise, which can affect diagnostic accuracy.

Purpose of the Study:

• To develop a novel denoising model for Magnetic Resonance (MR) images.
• To leverage compressive sensing with advanced regularization techniques.
• To enhance image quality by reducing noise while preserving crucial image details.

Main Methods:

• A new denoising model integrating compressive sensing with L1 and Hessian regularizations was proposed.
• L1 regularization was employed to promote sparsity and reduce noise in MR images.
• Hessian regularization was introduced to prevent the over-smoothing of image details.

Main Results:

• The proposed model effectively sparsifies the image, leading to significant noise reduction.
• Hessian regularization successfully protected fine details from being lost during the denoising process.
• Experimental results validated the model's efficiency and superior denoising capability compared to existing methods.

Conclusions:

• The developed compressive sensing-based denoising model offers an effective solution for MR image noise reduction.
• The combination of L1 and Hessian regularizations provides a balanced approach to noise removal and detail preservation.
• This method holds promise for improving the quality and diagnostic value of MR images.