Dual Residual Attention Network for ICMOS Sensing Image

Xia WANG; Xin ZHANG; Gangcheng JIAO; Ye YANG; Hongchang CHENG; Bo YAN

doi:10.3788/gzxb20225106.0610002

Acta Photonica Sinica, Volume. 51, Issue 6, 0610002(2022)

Dual Residual Attention Network for ICMOS Sensing Image

Xia WANG^1,2、*, Xin ZHANG², Gangcheng JIAO¹, Ye YANG¹, Hongchang CHENG¹, and Bo YAN¹

Author Affiliations

¹Science and Technology on Low-Light-Level Night Vision Laboratory，Xi'an 710065，China

²Key Laboratory of Optoelectronic Imaging Technology and System，Ministry of Education，School of Optics and Photonics，Beijing Institute of Technology，Beijing 100081，China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. The pipeline of ICMOS

Download full size

View in Article

Fig. 2. The overview of the method

Download full size

View in Article

Fig. 3. The structure of the residual block

Download full size

View in Article

Fig. 4. The structure of channel attention block

Download full size

View in Article

Fig. 5. Examples under different illumination in the datasets

Download full size

View in Article

Fig. 6. Denoising results of parrot image under $2 \times 10^{- 1}$ lx

Download full size

View in Article

Fig. 7. Denoising results of butterfly image under $2 \times 10^{- 1}$ lx

Download full size

View in Article

Fig. 8. Denoising results of parrot image under $3 \times 10^{- 2}$ lx

Download full size

View in Article

Fig. 9. Denoising results of butterfly image under $3 \times 10^{- 2}$ lx

Download full size

View in Article

Fig. 10. Denoising results of parrot image under $2 \times 10^{- 3}$ lx

Download full size

View in Article

Fig. 11. Denoising results of butterfly image under $2 \times 10^{- 3}$ lx

Download full size

View in Article

Table 1. Comparison of objective evaluation indicators of different methods

View table

View in Article

Table 1. Comparison of objective evaluation indicators of different methods

Methods	PSNR/dB（↑）				SSIM（↑）
Methods	$10^{- 1}$ lx	$10^{- 2}$ lx	$10^{- 3}$ lx	Average	$10^{- 1}$ lx	$10^{- 2}$ lx	$10^{- 3}$ lx	Average
Noisy images	24.17	20.39	18.41	20.99	0.543 9	0.409 2	0.355 0	0.436 0
NLM	26.97	21.82	19.54	22.78	0.712 0	0.503 7	0.431 6	0.549 1
DnCNN	24.26	20.45	18.45	21.05	0.549 5	0.413 3	0.357 8	0.440 2
CBDnet	24.36	20.50	18.51	21.12	0.555 6	0.417 4	0.363 2	0.445 4
VDN	23.21	21.23	20.00	21.48	0.565 1	0.449 0	0.408 0	0.474 0
BM3D	25.44	26.64	17.45	23.18	0.865 7	0.878 6	0.800 1	0.848 1
Ours	33.23	32.11	32.87	32.74	0.895 0	0.886 5	0.913 7	0.898 4

Table 2. The running time of different methods for processing a single 1 920×1 080 image
View table
View in Article
Table 2. The running time of different methods for processing a single 1 920×1 080 image
Methods Time cost/s Platform
NLM 564 Python（CPU）
DnCNN 0.179 Pytorch（GPU）
CBDnet 0.01 Pytorch（GPU）
VDN 0.14 Pytorch（GPU）
BM3D 1 868 Python（CPU）
Ours 0.005 Pytorch（GPU）

Tools

Get Citation

Copy Citation Text

Xia WANG, Xin ZHANG, Gangcheng JIAO, Ye YANG, Hongchang CHENG, Bo YAN. Dual Residual Attention Network for ICMOS Sensing Image[J]. Acta Photonica Sinica, 2022, 51(6): 0610002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites