基于低秩矩阵填充的MRI序列去噪

MRI sequence denoising based on low-rank matrix completion

现有的核磁共振成像(MRI)图像序列去噪方法大都针对单一的噪声类型,在实际中面对复杂的噪声情况其去噪效果并不理想。为改善这一现状,提出了一种基于低秩矩阵填充的MRI图像序列的去噪方法。针对混合噪声,通过在空间域和时域搜索和分组近似块,将去噪问题转换为低秩矩阵填充问题,使其不需要限定噪声的统计特性。实验结果表明,针对混合噪声,该算法的鲁棒性与去噪性能对比其他算法有所提升。

Most existing magnetic resonance imaging（ MRI） sequence denoising algorithms assume a single statistical model of image noise,which effect often is not ideal when facing complex noisy situation in practice. In order to improve this situation,this paper proposed a MRI sequence denoising method based on low-rank matrix. For mixed noise,by grouping similar patches in both spatial and temporal domain,it formulated the problem of removing mixed noise as a low-rank matrix completion problem,which leaded to a denoising scheme without assumptions on the statistical properties of noise. The experimental results show that in view of the mixed noise,the robustness and the denoising performance of the proposed denoising algorithm,have gain improve compared with other algorithms.