Fig. 1. Illustrative example of the multi-depth fast-denoising method
Fig. 2. Depth images and 3D point clouds of s of geometric shapes under different PPP and SBR. (a) Surfaces per pixel on geometric shapes; (b) Depth image estimated by MLE under PPP=11.12, SBR=3.40; (c)-(e) Depth images estimated by the proposed method respectively under PPP=11.12 and SBR=3.40, PPP=2.33 and SBR=0.99, PPP=1.08 and SBR=0.13, and (f)-(i) the corresponding 3D point clouds
Fig. 3. Depth images of the art scene. (a) Reference; (b)-(g) MLE, FDTCP, Rapp, SPISTA, ManiPOP, and proposed method, respectively
Fig. 4. 3D point clouds of the art scene. (a) Reference; (b)-(g) MLE, FDTCP, Rapp, SPISTA, ManiPOP, and proposed method, respectively
Fig. 5. Schematic diagram of long-range single-photon lidar system
Fig. 6. Complex buildings at 1 km. (a) Visible-band image; (b) Histogram of test dataset
Fig. 7. Depth images of complex buildings. (a) Reference; (b)-(g) MLE, FDTCP, Rapp, SPISTA, ManiPOP, and proposed method, respectively
Fig. 8. 3D point clouds of complex buildings. (a) Reference; (b)-(g) MLE, FDTCP, Rapp, SPISTA, ManiPOP, and proposed method, respectively
Method | RMSE/mm | SRE/dB | Processing time/s | MLE | 187.54 | 11.11 | 1.12 | FDTCP | 233.93 | 12.65 | 1.56 | Rapp | 196.41 | 12.52 | 154.43 | SPISTA | 172.14 | 16.67 | 197.12 | ManiPOP | 155.60 | 13.66 | 1103.26 | Proposed | 87.32 | 20.27 | 83.49 |
|
Table 1. RMSE, SRE and processing time of different methods on the art scene
Method | RMSE/m | SRE/dB | Processing time/s | MLE | 751.04 | 2.16 | 4.74 | FDTCP | 876.05 | 4.77 | 4.08 | Rapp | 424.68 | 15.81 | 161.75 | SPISTA | 65.86 | 30.12 | 553.88 | ManiPOP | 584.98 | 4.32 | 738.35 | Proposed | 48.04 | 35.66 | 133.30 |
|
Table 2. RMSE, SRE and processing time of different methods on complex buildings