6.2 The Theory RFID Multi-tag Image Denoising by FDnCNN
213
Fig. 6.9 FDnCNN denoising for different noise level a σ = 10 PSNR = 44.22; b σ = 30 PSNR
= 40.89; c σ = 50 PSNR = 37.82; d σ = 70 PSNR = 37.82
(3) For the noise grade σ σ[0,100], the noise level has little effect on the image
denoising results. Obviously, with the increase in noise level, clearer image
edges can be extracted, but the label’s details are weakened.
In Table 6.1, comparing FDnCNN with the current advanced methods (DnCNN
[18] and WNNM [17]), PSNR and SSIM are the averages of them in every multilabel images collected by CCD for 50 times. When σ = 20, 30, different denoising
techniques denoise multi-label images by a horizontal camera as shown in Fig. 6.10.
Referring to Table 6.1 and Fig. 6.10, it can be seen that.
Table 6.1 For different noise
level maps, the images
denoising quality evaluation
σ
WNNM
DnCNN
FDnCNN
10
43.0547/0.9850
42.1617/0.9718
45.0542/0.9803
20
38.6890/0.9668
42.2401/0.9736
42.7822/09,762
25
37.6780/0.9622
40.9402/0.9690
41.8047/0.9740
30
36.5203/0.9525
38.7932/0.9620
40.8968/0.9716
40
34.6586/0.9310
37.2450/0.9485
39.2790/0.9661