MR Image Restoration by Utilizing Fractional-Order TV and Recursive Filtering

Nana Wei, Wei Xue

Abstract: Total variation based methods are effective models in magnetic resonance image restoration. For eliminating impulse noise, an effective way is to use the l0-norm total variation model. However, the TV image restoration consistently produces staircase artifacts, particularly at noise levels with high density. In this paper, we propose a novel MR image restoration model incorporating fractional-order regularization and filtering methods. Specifically, the first term employs the l0-norm as the data fidelity term to effectively eliminate impulse noise. The second term introduces a fractional-order total variation regularizer, which preserves structural information while mitigating staircase artifacts during deblurring. Given its suboptimal performance in texture detail recovery, we utilize recursive filtering for high-quality edge-preserving filtering. Finally, we solve the corresponding optimization model by using the alternating direction method of multipliers. Experimental results demonstrate the effectiveness of our method in restoring medical images.

Submission Number: 148