A Weighted Two-Level Bregman Method with Dictionary Updating for Nonconvex MR Image Reconstruction

Figure 1

Performance of various algorithms at different undersampling ratios. (a) An axial T2-weighted reference image of the brain. ((b) and (c)) Pseudo radial sampling with 85% and 95% undersampling. (d) and (e) are the PSNR and HFEN versus the number of iterations of DLMRI, TBMDU, and WTBMDU. ((f), (g), (h), and (i)) The reconstruction results and the corresponding error magnitudes of DLMRI and WTBMDU-L1 with at 85% undersampling. ((j) and (h)) The reconstruction results of DLMRI and WTBMDU-L1 with at 95% undersampling.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)