Noise Reduction Method for Spectral Holographic Reconstructed Images Based on Noise Discreteness

Jiaxue He; Shuli Lou; Chao Lin

doi:10.3788/AOS232016

Acta Optica Sinica, Volume. 44, Issue 9, 0907001(2024)

Noise Reduction Method for Spectral Holographic Reconstructed Images Based on Noise Discreteness

Jiaxue He¹, Shuli Lou¹, and Chao Lin^2、*

Author Affiliations

¹School of Physics and Electronic Information, Yantai University, Yantai 264005, Shandong, China

²Aviation Operations and Service Institute, Naval Aviation University, Yantai 264000, Shandong, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(17)

Fig. 1. Optical cryptosystem framework

Download full size

Fig. 2. Partial results of optical cryptosystem in the digital simulation process. (a) Original image at 449 nm; (b) original image at 557 nm; (c) original image at 637 nm; (d) reconstructed image at 449 nm; (e) reconstructed image at 557 nm; (f) reconstructed image at 637 nm

Download full size

Fig. 3. Effect of noise intensity and type on C-PSNR of final color recovery image. (a) 3D surface plot of C-PSNR versus multiplicative noise amplitude and phase noise intensity; (b) 3D surface plot of C-PSNR versus additive noise amplitude and phase noise intensity; (c) effect of amplitude noise intensity on C-PSNR; (d) effect of phase noise intensity on C-PSNR

Download full size

Fig. 4. M-PSNR curve between reconstructed image and original image

Download full size

Fig. 5. Comparison of color recovery images obtained by DSA and WSA methods. (a)(b) Color recovery image obtained by DSA method and its local zoom-in; (c)(d) color recovery image obtained by WSA method and its local zoom-in

Download full size

Fig. 6. RGB three-component mean intensity distribution. (a) R-component mean intensity distribution; (b) G-component mean intensity distribution; (c) B-component mean intensity distribution

Download full size

Fig. 7. RGB three-wavelength center selection versus C-PSNR

Download full size

Fig. 8. Color recovery images under different wavelengths. (a) Color recovery image by reconstructed images at 621, 549, and 449 nm; (b) color recovery image by 89 wavelength reconstructed images

Download full size

Fig. 9. Comparison of interval radius versus C-PSNR or M-PSNR and final color recovery images. (a) Relationship between interval radius and C-PSNR using SRI+M-PSNR method; (b) relationship between interval radius and M-PSNR using DRI+M-PSNR method; (c) relationship between interval radius and C-PSNR using SRI+BW method; (d) relationship between interval radius and M-PSNR using DRI+BW method; (e) color recovery image obtained by SRI+M-PSNR method; (f) color recovery image obtained by DRI+M-PSNR method; (g) color recovery image obtained by SRI+BW method; (h) color recovery image obtained by DRI+BW method

Download full size

Fig. 10. Curve of interval radius versus C-PSNR. (a) Reconstructed image at an interval radius of 5 with C-PSNR of 42.25 dB; (b) reconstructed image at an interval radius of 26 with C-PSNR of 78.59 dB; (c) reconstructed image at an interval radius of 45 with C-PSNR of 74.12 dB

Download full size

Fig. 11. Threshold selection versus C-PSNR of the final color recovery image. (a) Reconstructed image with C-PSNR of 74.14 dB when threshold is 10 dB; (b) reconstructed image with C-PSNR of 78.27 dB when threshold is 30 dB; (c) reconstructed image with C-PSNR of 44.78 dB when threshold is 70 dB

Download full size

Fig. 12. Flowchart of WSA algorithm

Download full size

Fig. 13. Experimental results. (a) Reconstructed image of DSA; (b) reconstructed image of normalized M-PSNR as a weighting factor; (c) reconstructed image using only WSA algorithm; (d) reconstructed image using only CBM3D algorithm for 89 wavelengths of recovered images; (e) reconstructed image using WSA algorithm followed by the CBM3D algorithm; (f) reconstructed image using first BM3D algorithm followed by WSA algorithm

Download full size

Table 1. RGB three-component mean intensities and differences from the true values under different methods
View table
Table 1. RGB three-component mean intensities and differences from the true values under different methods
Analyzed object $I_{R}$ $I_{G}$ $I_{B}$ $(I_{R} - I_{O R}) ∶ (I_{G} - I_{O G}) ∶ (I_{B} - I_{O B})$
DSA 72.10 64.68 24.86 6.23∶16.14∶3.68
WSA 67.12 58.87 17.82 1.25∶10.34∶10.77
True value of average intensity 65.87 48.53 28.54 —

Table 2. Comparison of the ratio between mean intensities of RGB components in four methods
View table
Table 2. Comparison of the ratio between mean intensities of RGB components in four methods
Method $R_{r g}$ $R_{g b}$ $R_{r b}$
SRI+M-PSNR 1.06 1.58 1.67
DRI+M-PSNR 2.31 0.48 1.12
SRI+BW 1.07 1.32 1.41
DRI+BW 2.18 0.40 0.87
True value of average intensity 1.36 1.70 2.31

Table 3. Comparison of C-PSNR and wavelength number of four methods
View table
Table 3. Comparison of C-PSNR and wavelength number of four methods
Method C-PSNR /dB Number of wavelengths
DSA 74.13 89
WSA 64.63 89
IRA 78.59 70
TS 78.27 89

Table 4. Comparison of the use order of algorithms
View table
Table 4. Comparison of the use order of algorithms
Method C-PSNR /dB Processing time of each image /s
WSA 78.59 0.75
CBM3D 79.15 4.14
WSA+CBM3D 91.11 4.89
BM3D+WSA 88.69 3.04

Tools

Get Citation

Copy Citation Text

Jiaxue He, Shuli Lou, Chao Lin. Noise Reduction Method for Spectral Holographic Reconstructed Images Based on Noise Discreteness[J]. Acta Optica Sinica, 2024, 44(9): 0907001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites