Author Affiliations
Department of Electronic Engineering, Xi'an Aeronautical University, Xi'an, Shaanxi 710077, Chinashow less
Fig. 1. Comparison between proposed HA_OMP and OMP
Fig. 2. Original images of the 50th band. (a) Cuprite1; (b) Cuprite2; (c) Indian Pines; (d) Pavia University
Fig. 3. Influence of evolution generation, population size, and atom number on HA_OMP. (a) Atom number is 50; (b) population size is 10; (c) evolution generation is 5
Fig. 4. Reconstructed PSNR of Cuprite1 vs atom number. (a) OMP; (b) HA_OMP
Fig. 5. Comparison between two reconstructed Cuprite2 images. (a) Reconstructed image of OMP algorithm; (b) reconstructed image of HA_OMP algorithm
Fig. 6. Comparison between two reconstructed Pavia University images. (a) Reconstructed image of OMP algorithm; (b) reconstructed image of HA_OMP algorithm
Scene | Originalbands | Originalimage size | Availablebands | Croppedimage size |
---|
Cuprite1 | 224 | 614×512 | 188 | 256×256 | Cuprite2 | 224 | 614×512 | 188 | 256×256 | Indian Pines | 220 | 145×145 | 200 | 128×128 | Pavia University | 115 | 610×340 | 103 | 256×256 |
|
Table 1. Basic situation of four datasets
Scene | Algorithm | SR=0.1 | SR=0.2 | SR=0.3 | SR=0.4 | SR=0.5 |
---|
PSNR /dB | Atomnumber | PSNR /dB | Atomnumber | PSNR /dB | Atomnumber | PSNR /dB | Atomnumber | PSNR /dB | Atomnumber |
---|
Cuprite1 | OMP | 24.12 | 3 | 29.67 | 6 | 32.37 | 9 | 34.73 | 14 | 36.18 | 20 | HA_OMP | 24.76 | 5 | 28.76 | 6 | 31.90 | 21 | 34.75 | 33 | 36.76 | 50 | Cuprite2 | OMP | 21.88 | 3 | 27.22 | 5 | 29.12 | 8 | 31.04 | 14 | 32.05 | 18 | HA_OMP | 23.04 | 4 | 26.79 | 6 | 28.70 | 19 | 30.12 | 34 | 32.79 | 46 | Indian Pines | OMP | 13.01 | 3 | 16.75 | 3 | 17.85 | 6 | 19.09 | 11 | 20.34 | 18 | HA_OMP | 14.85 | 5 | 16.91 | 6 | 17.56 | 12 | 19.70 | 25 | 21.02 | 34 | PaviaUniversity | OMP | 19.78 | 2 | 22.03 | 4 | 23.54 | 7 | 25.38 | 14 | 27.02 | 22 | HA_OMP | 20.24 | 4 | 22.57 | 7 | 24.05 | 14 | 25.79 | 27 | 27.26 | 43 |
|
Table 2. Optimal PSNR for single band and required atom number
Scene | Algorithm | Time | SR=0.1 | SR=0.2 | SR=0.3 | SR=0.4 | SR=0.5 |
---|
Cuprite1 | OMP | Matching | 1031.98 | 1054.99 | 1049.73 | 1063.31 | 1086.11 | Updating | 9.65 | 11.36 | 12.13 | 11.57 | 12.50 | HA_OMP | Matching | 40.02 | 41.79 | 43.59 | 45.26 | 47.52 | Updating | 1.90 | 2.12 | 2.20 | 2.19 | 2.13 | Cuprite2 | OMP | Matching | 1056.40 | 1062.29 | 1079.52 | 1060.32 | 1090.40 | Updating | 11.30 | 12.72 | 12.11 | 14.66 | 12.65 | HA_OMP | Matching | 39.62 | 42.35 | 44.08 | 45.27 | 46.51 | Updating | 1.66 | 1.88 | 2.15 | 2.25 | 2.28 |
|
Table 3. Matching and updating time required for reconstructing single bands
Scene | Algorithm | SR=0.1 | SR=0.2 | SR=0.3 | SR=0.4 | SR=0.5 |
---|
Cuprite1 | OMP | 24.50 | 29.69 | 32.63 | 35.05 | 36.43 | HA_OMP | 24.54 | 28.96 | 33.06 | 34.70 | 37.19 | Cuprite2 | OMP | 21.96 | 27.49 | 29.42 | 31.34 | 32.52 | HA_OMP | 22.68 | 27.33 | 29.30 | 32.09 | 33.15 | Indian Pines | OMP | 14.94 | 18.86 | 20.27 | 21.79 | 22.93 | HA_OMP | 17.09 | 19.21 | 21.51 | 21.90 | 23.53 | Pavia University | OMP | 19.22 | 21.85 | 23.36 | 24.99 | 26.48 | HA_OMP | 20.64 | 22.38 | 23.74 | 25.43 | 26.59 |
|
Table 4. Average optimal PSNR of two algorithms under different sampling ratesdB
Scene | Algorithm | SR=0.1 | SR=0.2 | SR=0.3 | SR=0.4 | SR=0.5 |
---|
Cuprite1 | OMP | 1 | 1 | 1 | 1 | 1 | HA_OMP | 15.10 | 8.62 | 12.39 | 12.71 | 11.34 | Cuprite2 | OMP | 1 | 1 | 1 | 1 | 1 | HA_OMP | 15.71 | 9.59 | 9.80 | 10.88 | 10.42 | Indian Pines | OMP | 1 | 1 | 1 | 1 | 1 | HA_OMP | 17.67 | 14.11 | 9.58 | 11.82 | 11.19 | Pavia University | OMP | 1 | 1 | 1 | 1 | 1 | HA_OMP | 15.53 | 13.99 | 10.80 | 10.63 | 12.18 |
|
Table 5. Acceleration times of HA_OMP relative to OMP under different sampling rates