Super-Resolution of Magnetic Resonance Images via Convex Optimization with Local and Global Prior Regularization and Spectrum Fitting

Figure 10

Differences in PSNR performance of the proposed method when the boundary became redundant (positive values) or insufficient (negative values) compared with the true boundary (the value is zero). The input image of the left figure was T1-weighted image of normal brain phantom, and that of the right one was T2-weighted image of lesional brain phantom.