Phase Retrieval Algorithm Fusing Multiple Wavelets and Total Variation Regularization

Qiusheng Lian; Ying Li; Shuzhen Chen

doi:10.3788/AOS201838.0220001

Journals >Acta Optica Sinica >Volume 38 >Issue 2 >Page 0220001 > Article

Acta Optica Sinica
Vol. 38, Issue 2, 0220001 (2018)

Phase Retrieval Algorithm Fusing Multiple Wavelets and Total Variation Regularization

Qiusheng Lian^*, Ying Li, and Shuzhen Chen

Author Affiliations

Institute of Information Science and Technology, Yanshan University, Qinhuangdao, Hebei 066004, China

show less

DOI: 10.3788/AOS201838.0220001 Cite this Article Set citation alerts

Qiusheng Lian, Ying Li, Shuzhen Chen. Phase Retrieval Algorithm Fusing Multiple Wavelets and Total Variation Regularization[J]. Acta Optica Sinica, 2018, 38(2): 0220001 Copy Citation Text

show less

Fig. 1. Wavelet bases of wavelet transform. (a) db10; (b) sym4

Download full size

Comparison of phase retrieval results (PSNR,SSIM) of different algorithms whenRSNR=15 dB. (a) WF (13.28 dB,0.08); (b) DOLPHIn (27.13 dB,0.65); (c) BM3D-PRGAMP (29.13 dB, 0.80); (d) PRWATV (30.82 dB, 0.83)

Fig. 2. Comparison of phase retrieval results (PSNR,SSIM) of different algorithms when RSNR =15 dB. (a) WF (13.28 dB,0.08); (b) DOLPHIn (27.13 dB,0.65); (c) BM3D-PRGAMP (29.13 dB, 0.80); (d) PRWATV (30.82 dB, 0.83)

Download full size

Fig. 3. Partial information comparison of phase retrieval results of different algorithms when RSNR =15 dB. (a) WF; (b) DOLPHIn; (c) BM3D-PRGAMP; (d) PRWATV

Download full size

Fig. 4. Comparison of phase retrieval results (PSNR, SSIM) of different algorithms when RSNR =15 dB. (a) TWF (13.10 dB, 0.11); (b) DOLPHIn (25.91 dB, 0.62); (c) SPAR (28.88 dB, 0.77); (d) PRWATV (28.68 dB, 0.76)

Download full size

Threshold algorithm	PSNR
Threshold algorithm	R_SNR =20 dB	R_SNR =15 dB	R_SNR =10 dB
Soft	28.72	26.55	24.16
Hard	29.67	27.08	24.34
Group soft	29.34	27.13	24.74
Group hard	30.57	28.05	25.40

Table 1. Comparison of PSNR obtained by phase retrieval experiments with different threshold algorithms dB

Regularization term	PSNR
Regularization term	R_SNR =20 dB	R_SNR =15 dB	R_SNR =10 dB
db10 sparsity	28.45	25.71	22.97
sym4 sparsity	28.21	25.63	22.99
Total variation	28.97	27.01	24.77
db10 sparsity+ total variation	30.17	27.72	25.15
sym4 sparsity+ total variation	30.04	27.67	25.13
db10 sparsity+ sym4 sparsity	30.06	27.36	24.67
db10 sparsity+ sym4 sparsity+ total variation	30.57	28.05	25.40

Table 2. Comparison of PSNR obtained by phase retrieval experiments with different regularization termsdB

Image	WF	DOLPHIn	BM3D-PRGAMP	PRWATV
Lena512	13.29	27.13	29.13	30.82
Boat512	13.16	25.38	27.53	28.39
Mandril512	13.73	20.40	25.42	26.36
Barbara512	12.89	25.10	26.60	26.64
Fingerprint512	12.49	22.50	24.85	25.74
Cameraman256	12.15	24.51	27.00	27.54
Peppers256	12.94	24.64	27.71	28.61
House256	12.92	25.24	29.88	30.05

Table 3. Comparison of PSNR obtained by phase retrieval experiments with RSNR =15 dBdB

R_SNR /dB	Algorithm	256 pixel×256 pixel	512 pixel×512 pixel
R_SNR /dB	Algorithm	PSNR /dB		Time /s	PSNR /dB	Time /s
20	WF	13.49	0.96	14.00	6.21
	DOLPHIn	27.82	11.05	27.17	33.71
	BM3D-PRGAMP	30.52	22.26	29.10	98.66
	PRWATV	31.38	1.98	29.90	12.65
15	WF	12.67	0.91	13.11	6.07
	DOLPHIn	27.82	9.95	24.10	29.02
	BM3D-PRGAMP	28.20	47.89	26.70	197.04
	PRWATV	28.73	2.12	27.59	12.58

Table 4. Comparison of average PSNR and running time obtained by phase retrieval experiments with different noise measurement values

R_SNR /dB	Algorithm									Lena512		Boat512		Cameraman256		Peppers256
R_SNR /dB										PSNR /dB	Time /s	PSNR /dB	Time /s	PSNR /dB	Time /s	PSNR /dB	Time /s
20										TWF	14.10	7.37	14.06	7.21	12.71	1.76	13.80	1.72
										DOLPHIn	29.38	21.00	27.52	20.84	26.55	8.01	27.55	7.85
										SPAR	34.19	100.20	31.64	101.29	30.98	25.38	32.24	25.40
										PRWATV	33.60	11.12	31.31	11.04	30.71	1.92	31.89	1.92
15										TWF	13.34	7.58	13.09	7.38	12.13	1.76	12.88	1.72
										DOLPHIn	27.43	28.55	25.90	28.57	24.65	7.32	25.53	7.20
	SPAR	31.37	101.38	28.88	100.92	27.85	25.87	29.12	25.77
	PRWATV	30.97	11.24	28.68	11.34	27.82	1.66	28.79	1.66

Table 5. Comparison of PSNR and running time obtained by phase retrieval experiments with different noise

Qiusheng Lian, Ying Li, Shuzhen Chen. Phase Retrieval Algorithm Fusing Multiple Wavelets and Total Variation Regularization[J]. Acta Optica Sinica, 2018, 38(2): 0220001

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information