Undersampled MR Image Reconstruction with Data-Driven Tight Frame

Figure 5

(a) Original image. (b) Radial sampling mask. (c) PSNR versus noise standard deviation. The error magnitudes of reconstructed results from the 3-fold radial sampling mask with noise standard deviation 4 by (d) CSMRI-TV, (e) DLMRI, and (f) our proposed DDTF, respectively.

(a)

(b)

(c)

(d)

(e)

(f)