Compressed Sensing MRI Reconstruction from Highly Undersampled -Space Data Using Nonsubsampled Shearlet Transform Sparsity Prior

Figure 10

Performance with noise standard deviation 10.2 for a water phantom. (a) The noisy fully sampled reconstruction, (b) added noise magnitude, and (c)~(g) reconstructed images using zero-filling, TV-CG, ODWT-based, SFLCT-based, and the proposed NSST-based CS-MRI method, respectively. The PSNRs of (c)~(g) are 23.76, 31.31, 28.64, 30.34, and 32.52, respectively. (h)~(l) The difference images between fully sampled MR image and reconstruction images (c)~(g), respectively. The RLNEs of them are 0.1364, 0.0589, 0.0778, 0.0639, and 0.0498, respectively.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)