The method based on L1 norm optimization model for stripe noise removal of remote sensing image

Kai LI; Wen-Li LI; Chang-Pei HAN

doi:10.11972/j.issn.1001-9014.2021.02.018

Journals >Journal of Infrared and Millimeter Waves >Volume 40 >Issue 2 >Page 272 > Article

Journal of Infrared and Millimeter Waves
Vol. 40, Issue 2, 272 (2021)

The method based on L1 norm optimization model for stripe noise removal of remote sensing image

Kai LI^1、2、3, Wen-Li LI^1、2、3, and Chang-Pei HAN^1、2、*

Author Affiliations

¹Shanghai Institute of Technical Physics，Chinese Academy of Sciences，Shanghai 200083，China

²Key Laboratory of Infrared Detection and Imaging Technology，Chinese Academy of Sciences，Shanghai 200083，China

³University of Chinese Academy of Sciences，Beijing 100049，China

show less

DOI: 10.11972/j.issn.1001-9014.2021.02.018 Cite this Article

Kai LI, Wen-Li LI, Chang-Pei HAN. The method based on L1 norm optimization model for stripe noise removal of remote sensing image[J]. Journal of Infrared and Millimeter Waves, 2021, 40(2): 272 Copy Citation Text

show less

Fig. 1. The framework of the proposed model

Download full size | View in the Article

(a) The original remote sensing image, (b) weighting factor image in Eq. 8, (c) the smooth part, (d) the high frequency part, (e) weighting factor image in Eq. 9, (f) edge weighting image in Eq. 10

Fig. 2. (a) The original remote sensing image, (b) weighting factor image in Eq. 8, (c) the smooth part, (d) the high frequency part, (e) weighting factor image in Eq. 9, (f) edge weighting image in Eq. 10

Download full size | View in the Article

Fig. 3. Destriped results of AGRI band 11 subimage (a) original image, (b) WFAF, (c) SLD, (d) UTV, (e) proposed method

Download full size | View in the Article

Fig. 4. Mean line profiles for images shown in Fig. 3, (a) WFAF, (b) SLD, (c) UTV, (d) proposed method

Download full size | View in the Article

Fig. 5. Column-averaged power spectrum for images shown in Fig. 3 (a) original image, (b) WFAF, (c) SLD, (d) UTV, (e) proposed method

Download full size | View in the Article

Fig. 6. Destriped results of AGRI band 11 subimage, (a) original image, (b) WFAF, (c) SLD, (d) UTV, (e) proposed method

Download full size | View in the Article

Fig. 7. The extracted stripe components of different algorithms (a) WFAF, (b) SLD, (c) UTV, (d) proposed method

Download full size | View in the Article

Fig. 8. Mean line profiles for images shown in Fig. 6 (a) WFAF, (b) SLD, (c) UTV, (d) proposed method

Download full size | View in the Article

Fig. 9. Column-averaged power spectrum for images shown in Fig. 6, (a) original image, (b) WFAF, (c) SLD, (d) UTV, (e) proposed method

Download full size | View in the Article

Fig. 10. (a) Reference image, (b) simulated stripe image

Download full size | View in the Article

Fig. 11. (a) The PSNR curve with

λ_{1}

as independent variable, (b) The PSNR curve with

λ_{2}

as independent variable

Download full size | View in the Article

Fig. 12. Processing results of different parameters

λ_{1}

Download full size | View in the Article

Fig. 13. Processing results of different parameters

λ_{2}

Download full size | View in the Article

Fig. 14. Destriping result of AGRI band 9 images with the proposed algorithm

Download full size | View in the Article

Fig. 15. Destriping result of AGRI band 10 images with the proposed algorithm

Download full size | View in the Article

Fig. 16. Destriping result of AGRI band 14 images with the proposed algorithm

Download full size | View in the Article

1：Input：Stripe image f，parameters $λ_{1}$ ， $λ_{2}$ ， $β_{1}$ ， $β_{2}$ ， $β_{3}$ ， $δ$ and $S$ .
2：Initialize：Set $s^{0} = 0$ ， $Z^{0} = V^{0} = 0$ ， $H^{0} = \nabla_{y} f$ ， $p_{1} = 0$ ， $p_{2} = 0$ ， $p_{3} = 0$ ，and $ε = 10^{- 4}$ .
3：Solve $W_{f}$ by（10）
4：While $‖(f - s^{k}) - (f - s^{k - 1})‖ / ‖f - s^{k}‖ > ε$ and $k < N_{m a x}$ do
5：Solve $Z^{k + 1}$ ， $V^{k + 1}$ ， $H^{k + 1}$ using a thresholding method by（14），（17），（19）
6：Solve $s^{k + 1}$ using FFT by（21）
7：Update $p_{1}^{k + 1}$ ， $p_{2}^{k + 1}$ ，and $p_{3}^{k + 1}$ by（22）
8： End while
9：Output： $u^{k + 1} = f - s^{k + 1}$ .

Table 1. The proposed destriping algorithm

View in the Article

No.	Central Band /μm	Spectral Band /μm	Spatial Resolution	Number of pixels	Main Application
9	6.25	5.80∼6.70	4 km	4*1	upper-level water vapor
10	7.10	6.90∼7.30	4 km	4*1	mid-level water vapor
11	8.50	8.00∼9.00	4 km	4*1	integrated water vapor，cloud
14	13.50	13.20∼13.80	4 km	4*1	cloud，water vapor

Table 2. Spectral parameters

View in the Article

Image	Index	WFAF	SLD	UTV	Proposed
AGRI band 11 Periodical stripes noise	NR	10.06	10.78	12.93	13.67
	MRD（%）	1.781 7	0.734 2	4.033 4	0.875 1
	ID	0.987 8	0.999 9	0.857 0	0.998 4
AGRI band 14 Random stripe noise	NR	7.830 2	6.400 5	5.316 5	8.365 9
	MRD（%）	3.086 2	3.303 6	4.778 2	3.065 3
	ID	0.945 5	0.999 9	0.985 1	0.998 8

Table 3. Qualitative results using NR, MRD and ID

Kai LI, Wen-Li LI, Chang-Pei HAN. The method based on L1 norm optimization model for stripe noise removal of remote sensing image[J]. Journal of Infrared and Millimeter Waves, 2021, 40(2): 272

Download Citation

Tools

Save the article for my favorites

Paper Information