基于奇异值分解的压缩感知核磁共振图像重构算法

针对传统压缩感知在核磁共振成像中存在着重构算法慢、成像时间长的缺点,利用核磁共振图像自身非满秩的特点,在压缩感知框架下以奇异值分解作为基底对图像稀疏表示进行了研究,并对重构算法进行了优化。实验结果表明,提出的奇异值方法在重构效果上能达到与小波稀疏变换法相近的峰值性噪比,且能有效缩短图像重构时间,达到加速核磁共振成像的目的。

Single value decomposition based compressed sensingMRI reconstruction algorithm

Traditional compressed sensing in magnetic resonance(MR) imaging has the shortcomings of slow reconstruction algorithm and long time imaging. This paper applied the non-full rank of MR image, and presented singular value decomposition (SVD) as sparse representation in the compressed sensing frame. At the same time, optimizing the reconstruction algorithm enhanced this theory. The experimental results show that the proposed SVD algorithm can reach similar reconstruction quality as wavelet based method in terms of peak signal to noise ratio. Meanwhile, it can effectively reduce reconstruction time and receive the object of fastening MR imaging.