A Dictionary Learning Method with Total Generalized Variation for MRI Reconstruction

Figure 2

Reconstruction of an axial T2-weighed brain image at 7–11-fold undersampling. (a) The reference image. ((b), (c)) The PSNR and HFEN versus the number of iterations. ((d), (e), (f)) The reconstruction results under 2D random sampling by three methods DLMRI, SHTGV, and DLTGV. ((g), (h), (i)) The corresponding reconstruction error magnitudes of (d), (e), and (f).

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)