Author Affiliations
1Engineering Research Center of Industrial CT Nondestructive Testing, Ministry of Education, Chongqing University, Chongqing 400044, China2College of Optoelectronic Engineering, Chongqing University, Chongqing 400044, Chinashow less
Fig. 1. Relationship diagram of similar window and search window of NLM algorithm
Fig. 2. Structural tensor decomposition. (a) Original image; (b) schematic of structure tensor
Fig. 3. Flowchart of ST-NLM algorithm
Fig. 4. Gaussian noise simulated image and local magnification of filtering results at different noise levels. (a) Simulated image; (b) σ=1; (c) σ=4; (d) σ=5; (e) σ=12
Fig. 5. CT image of spatial resolution testing card. (a) Original image; (b) NLM; (c) method in Ref. [12]; (d) ST-NLM
Fig. 6. Gray curves obtained by different filtering methods
Fig. 7. Typical CT slices of insect 1. (a) Original image; (b) locally enlarged image; (c) NLM; (d) method in Ref. [12]; (e) ST-NLM
Fig. 8. CT image of the 520th slice of insect 1. (a) Original image; (b) locally enlarged image; (c) NLM; (d) method in Ref. [12]; (e) ST-NLM
Fig. 9. CT image of the 434th slice of insect 2. (a) Original image; (b) locally enlarged image; (c) NLM; (d) method in Ref. [12]; (e) ST-NLM
Evaluation parameter | σ | Noisy image | NLM | Method in Ref. [12] | ST-NLM |
---|
| 1 | 0.2771 | 0.9444 | 0.9553 | 0.9556 | SSIM | 4 | 0.1987 | 0.8472 | 0.9025 | 0.9101 | | 5 | 0.1629 | 0.7997 | 0.8241 | 0.8663 | | 12 | 0.1150 | 0.6276 | 0.6722 | 0.7309 | | 1 | 23.54 | 30.86 | 32.04 | 36.39 | RPSN /dB | 4 | 21.84 | 30.11 | 31.17 | 33.35 | | 5 | 18.64 | 28.03 | 29.06 | 29.67 | | 12 | 15.31 | 24.19 | 25.01 | 25.08 |
|
Table 1. Quantitative evaluation of SSIM and PSNR by different methods
Object | CT system | X-ray energy /keV | Detector size /mm | Exposure time /ms | Views of projection |
---|
Spatial resolution testing card | Linear array | 4000 | 1.500 | 20 | 2048 | Insect 1 | Planer array | 150 | 0.200 | 500 | 500 | Insect 2 | Planer array | 60 | 0.075 | 1000 | 500 |
|
Table 2. CT system parameters for experiment
Image | Image size | Similar window size | Search window size | σ |
---|
Fig. 5 | 336×283 | 3×3 | 7×7 | 0.34 | | 768×1024 | 3×3 | 7×7 | 0.11 | Fig. 7 | 768×1024 | 3×3 | 7×7 | 0.14 | | 768×1024 | 3×3 | 7×7 | 0.12 | | 768×1024 | 3×3 | 7×7 | 0.10 | Fig. 8 | 768×1024 | 3×3 | 7×7 | 0.15 | Fig. 9 | 1200×1200 | 3×3 | 7×7 | 0.25 |
|
Table 3. Experimental parameters of different images
Image | Evaluation parameter | Slice No. | NLM | Method in Ref. [12] | ST-NLM |
---|
| | 340th | 0.0868 | 0.0941 | 0.0995 | Fig. 7 | Global Tenengrad | 592th | 0.1127 | 0.1209 | 0.1416 | | | 636th | 0.1063 | 0.1135 | 0.1355 | | | 710th | 0.0906 | 0.0978 | 0.0966 | Fig. 8 | Global Tenengrad | 520th | 0.0882 | 0.0902 | 0.1021 | Fig. 9 | Global Tenengrad | 434th | 0.0911 | 0.0944 | 0.1113 |
|
Table 4. Index of image sharpness
Image | Image size | NLM | NLM integrating image acceleration | Method in Ref. [12] | ST-NLM |
---|
Fig. 5 | 336×283 | 3.18 | 0.53 | 3.21 | 2.05 | Fig. 8 | 768×1024 | 20.10 | 1.78 | 21.10 | 8.06 | Fig. 9 | 1200×1200 | 37.90 | 3.86 | 38.30 | 15.30 |
|
Table 5. Comparison of operating time among various filtering algorithmss