Research Article
Sparse Dictionary-Based Magnetic Resonance Superresolution Imaging with Joint Loss Function Learning
Table 4
Comparison of superresolution PSNR for different block and overlap block training dictionaries (superresolution ratio 1 : 4).
| Block size | Overlapped block | Sample set 1 | Sample set 2 | Sample set 3 | Sample set 4 | Sample set 5 |
| 3 × 3 | 2 | 29.22/0.8619 | 25.05/0.7746 | 28.78/0.8607 | 26.91/0.8014 | 27.11/0.8273 | 4 × 4 | 3 | 29.45/0.8677 | 25.16/0.7832 | 28.93/0.8704 | 26.97/0.8085 | 27.26/0.8344 | 5 × 5 | 4 | 29.82/0.8755 | 25.34/0.7913 | 29.16/0.8758 | 27.09/0.8172 | 27.54/0.8401 | 6 × 6 | 5 | 30.19/0.8824 | 25.57/0.7988 | 29.43/0.8812 | 27.26/0.8200 | 27.81/0.8503 | 7 × 7 | 6 | 30.63/0.8875 | 25.72/0.8104 | 29.68/0.8875 | 27.30/0.8276 | 28.06/0.8556 | 8 × 8 | 7 | 30.80/0.8908 | 25.88/0.8233 | 29.96/0.8908 | 27.35/0.8302 | 28.25/0.8623 | 9 × 9 | 8 | 30.92/0.8994 | 25.94/0.8399 | 30.03/0.8966 | 27.43/0.8483 | 28.30/0.8701 | 10 × 10 | 9 | 31.11/0.9109 | 25.96/0.8487 | 30.15/0.9043 | 27.44/0.8514 | 28.31/0.8722 | 11 × 11 | 10 | 30.91/0.9025 | 25.91/0.8410 | 30.09/0.9002 | 27.32/0.8311 | 28.22/0.8614 | 12 × 12 | 11 | 30.87/0.8902 | 25.90/0.8315 | 30.05/0.8954 | 27.30/0.8292 | 28.17/0.8600 | 13 × 13 | 12 | 30.63/0.8843 | 25.76/0.8184 | 29.88/0.8871 | 27.19/0.8145 | 27.98/0.8496 |
|
|