Sparse Image Reconstruction Based on Improved Total Generalized Variation

Xiaozheng Ban; Zhihua Li; Beibei Li; Minda Xu

doi:10.3788/LOP55.111103

Journals >Laser & Optoelectronics Progress >Volume 55 >Issue 11 >Page 111103 > Article

Laser & Optoelectronics Progress
Vol. 55, Issue 11, 111103 (2018)

Sparse Image Reconstruction Based on Improved Total Generalized Variation

Xiaozheng Ban, Zhihua Li^*, Beibei Li, and Minda Xu

Author Affiliations

School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China

show less

DOI: 10.3788/LOP55.111103 Cite this Article Set citation alerts

Xiaozheng Ban, Zhihua Li, Beibei Li, Minda Xu. Sparse Image Reconstruction Based on Improved Total Generalized Variation[J]. Laser & Optoelectronics Progress, 2018, 55(11): 111103 Copy Citation Text

show less

Fig. 1. Eight neighborhoods of pixel points

Download full size

Fig. 2. Image data used in the test. (a) Sheep-Logan image; (b) Forbild-Head image; (c) Forbild-Abdomen image

Download full size

Fig. 3. ROI regions of the test images. (a) Sheep-Logan image; (b) Forbild-Head image; (c) Forbild-Abdomen image

Download full size

Fig. 4. Reconstructed image and ROI comparison of different algorithms in finite angle (Sheep-Logan). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 5. Reconstructed image and ROI comparison of different algorithms in finite angle (Forbild-Head). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 6. Reconstructed image and ROI comparison of different algorithms in finite angle (Forbild-Abdomen). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 7. Reconstructed image and ROI comparison of different algorithms in sparse angle (Sheep-Logan). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 8. Reconstructed image and ROI comparison of different algorithms in sparse angle (Forbild-Head). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 9. Reconstructed image and ROI comparison of different algorithms in sparse angle (Forbild-Abdomen). (a) Original image; (b) L1-Bregman algorithm; (c) TV-ADM algorithm; (d) TVAL3 algorithm; (e) TGV-ADM algorithm; (f) proposed algorithm

Download full size

Fig. 10. Comparison of vertical cross-sectional views of 135th column of each reconstruction image. (a) Sheep-Logan image; (b) Forbild-Head image; (c) Forbild-Abdomen image

Download full size

Parameter	β	μ_i(i=1,2,3)	γ	α₀	α₁
Value	64	32	1	0.2	2

Table 1. Initialization parameters

Algorithm	Image	SSIM	PSNR /dB	RMSE	Time /s
L1-Bregman	Sheep-Logan	0.549	15.748	0.163	10.969
	Forbild-Head	0.430	13.175	0.219	11.063
	Forbild-Abdomen	0.535	15.672	0.165	11.328
TV-ADM	Sheep-Logan	0.891	22.203	0.078	16.467
	Forbild-Head	0.893	21.916	0.080	16.453
	Forbild-Abdomen	0.967	33.024	0.022	15.781
TVAL3	Sheep-Logan	0.894	22.807	0.073	19.625
	Forbild-Head	0.899	23.072	0.070	20.109
	Forbild-Abdomen	0.968	33.643	0.021	22.234
TGV-ADM	Sheep-Logan	0.899	25.039	0.056	70.016
	Forbild-Head	0.935	27.420	0.043	69.594
	Forbild-Abdomen	0.967	35.312	0.018	68.609
Proposed	Sheep-Logan	0.903	26.091	0.054	150.682
	Forbild-Head	0.954	29.229	0.035	149.531
	Forbild-Abdomen	0.977	39.231	0.011	147.000

Table 2. Comparison of objective evaluation parameters of reconstructed images of various algorithms under finite projection angle

Algorithm	Image	SSIM	PSNR /dB	RMSE	Time /s
L1-Bregman	Sheep-Logan	0.619	17.449	0.134	11.625
	Forbild-Head	0.185	14.441	0.190	11.078
	Forbild-Abdomen	0.543	16.340	0.152	12.953
TV-ADM	Sheep-Logan	0.995	45.429	0.005	16.797
	Forbild-Head	0.993	39.136	0.011	16.531
	Forbild-Abdomen	0.981	36.339	0.015	17.250
TVAL3	Sheep-Logan	0.995	46.075	0.005	20.953
	Forbild-Head	0.995	43.074	0.007	20.234
	Forbild-Abdomen	0.985	38.522	0.012	22.000
TGV-ADM	Sheep-Logan	0.991	44.582	0.006	69.031
	Forbild-Head	0.995	43.810	0.006	67.406
	Forbild-Abdomen	0.978	39.608	0.011	71.922
Proposed	Sheep-Logan	0.999	50.018	0.003	144.250
	Forbild-Head	0.999	52.172	0.003	146.359
	Forbild-Abdomen	0.991	39.930	0.010	149.141

Table 3. Comparison of objective evaluation parameters of reconstructed images of various algorithms under sparse projection angle

Xiaozheng Ban, Zhihua Li, Beibei Li, Minda Xu. Sparse Image Reconstruction Based on Improved Total Generalized Variation[J]. Laser & Optoelectronics Progress, 2018, 55(11): 111103

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information