Super-Resolution Reconstruction Algorithm Based on Adaptive Image Online Dictionary Learning

Deqiang Cheng; Wenjie Yu; Xin Guo; Huandong Zhuang; Xinzhu Fu

doi:10.3788/LOP57.061505

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 6 >Page 061505 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 6, 061505 (2020)

Super-Resolution Reconstruction Algorithm Based on Adaptive Image Online Dictionary Learning

Deqiang Cheng^*, Wenjie Yu^**, Xin Guo, Huandong Zhuang, and Xinzhu Fu

Author Affiliations

School of Information and Control Engineering, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China

show less

DOI: 10.3788/LOP57.061505 Cite this Article Set citation alerts

Deqiang Cheng, Wenjie Yu, Xin Guo, Huandong Zhuang, Xinzhu Fu. Super-Resolution Reconstruction Algorithm Based on Adaptive Image Online Dictionary Learning[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061505 Copy Citation Text

show less

Fig. 1. Super-resolution reconstruction based on sparse encoding

Download full size

Fig. 2. Visualization of adaptive value for the image of Lenna

Download full size

Fig. 3. Comparison of reconstruction for different images when the magnification factor is 2. (a) Baby; (b Lenna; (c) butterfly

Download full size

Fig. 4. Comparison of reconstruction for image of baby when the magnification factor is 3

Download full size

Fig. 5. Comparison of reconstruction for image of baby when the magnification factor is 4

Download full size

Fig. 6. Comparison of reconstruction under Gaussian noise environment with variance of 0.0005

Download full size

Fig. 7. Comparison of reconstruction under pepper & salt noise environment with density of 0.001

Download full size

Method	Evaluation index	Baby	Lenna	Butterfly	Subway	Man	Bike	Baboon	Monarch
	RMSE	4.157	5.455	12.572	10.009	10.213	15.479	17.337	6.673
Bicubic	PSNR	37.059	34.713	27.462	29.440	29.261	25.657	24.674	32.958
	SSIM	0.951	0.911	0.915	0.871	0.845	0.850	0.696	0.959
	RMSE	4.622	12.108	15.991	11.624	10.926	21.070	19.442	7.543
NE	PSNR	36.889	28.669	26.559	27.416	28.291	21.657	22.589	31.211
	SSIM	0.893	0.788	0.737	0.753	0.786	0.591	0.505	0.882
	RMSE	3.601	4.561	7.836	8.623	8.863	11.357	16.157	4.701
Ours1	PSNR	38.384	36.299	31.572	30.802	30.500	28.401	25.286	36.031
	SSIM	0.963	0.927	0.958	0.908	0.880	0.914	0.759	0.970
	RMSE	3.603	4.574	7.845	8.520	8.881	11.486	16.141	4.718
Ours2	PSNR	38.351	36.276	31.562	30.846	30.483	28.322	25.295	35.996
	SSIM	0.963	0.926	0.958	0.909	0.880	0.9127	0.759	0.971
	RMSE	3.605	4.571	7.843	8.564	8.854	11.414	16.148	4.675
Ours3	PSNR	38.360	36.280	31.563	30.850	30.508	28.353	25.291	36.079
	SSIM	0.963	0.926	0.958	0.910	0.880	0.914	0.760	0.971

Table 1. Numerical comparison of reconstruction for different images when the magnification factor is 2

Method	Evaluation index	Baby	Lenna	Butterfly	Subway	Man	Bike	Baboon	Monarch
	RMSE	3.656	4.631	7.911	8.696	8.885	11.619	16.218	5.145
ScSr1^[14]	PSNR	38.262	36.215	31.502	30.707	30.484	28.207	25.269	35.300
	SSIM	0.961	0.925	0.956	0.906	0.878	0.910	0.757	0.967
	RMSE	3.633	4.593	7.884	8.672	8.866	11.594	16.185	5.104
ODL1^[17]	PSNR	38.270	36.239	31.518	30.738	30.497	28.226	25.272	35.309
	SSIM	0.962	0.926	0.957	0.907	0.879	0.911	0.758	0.968
	RMSE	3.601	4.561	7.836	8.623	8.863	11.357	16.157	4.701
Ours1	PSNR	38.384	36.299	31.572	30.802	30.500	28.401	25.286	36.031
	SSIM	0.963	0.927	0.958	0.908	0.880	0.914	0.759	0.97
	RMSE	3.693	4.692	8.268	8.802	8.995	12.295	16.173	4.881
ScSr2^[14]	PSNR	38.205	36.083	31.158	30.571	30.377	27.796	25.277	35.762
	SSIM	0.961	0.921	0.955	0.903	0.876	0.903	0.757	0.969
	RMSE	3.651	4.652	8.108	8.776	8.986	12.282	16.167	4.834
ODL2^[17]	PSNR	38.231	36.127	31.274	30.592	30.385	27.801	25.281	35.785
	SSIM	0.962	0.925	0.956	0.905	0.877	0.904	0.758	0.970
	RMSE	3.603	4.574	7.845	8.520	8.881	11.486	16.141	4.718
Ours2	PSNR	38.351	36.296	31.562	30.846	30.483	28.322	25.295	35.996
	SSIM	0.963	0.926	0.958	0.909	0.880	0.9127	0.759	0.971
	RMSE	3.751	4.801	9.309	8.982	9.385	12.963	16.179	5.479
ScSr3^[14]	PSNR	38.080	35.887	30.043	30.415	30.134	27.206	25.283	34.791
	SSIM	0.960	0.923	0.942	0.901	0.874	0.895	0.758	0.966
	RMSE	3.707	4.768	9.235	8.966	9.220	12.944	16.154	5.406
ODL3^[17]	PSNR	38.106	35.912	30.145	30.440	30.157	27.214	25.288	34.810
	SSIM	0.961	0.924	0.948	0.902	0.875	0.896	0.759	0.967
	RMSE	3.605	4.571	7.843	8.564	8.854	11.414	16.148	4.675
Ours3	PSNR	38.360	36.290	31.563	30.850	30.508	28.353	25.291	36.079
	SSIM	0.963	0.926	0.958	0.910	0.880	0.914	0.760	0.971

Table 2. Numerical comparison of reconstruction for different images when the magnification factor is 2

Method	Evaluation index	Baby	Man	Butterfly	Baboon
	RMSE	5.956	13.206	18.611	20.992
Bicubic	PSNR	33.941	27.032	24.056	23.013
	SSIM	0.904	0.749	0.819	0.543
	RMSE	6.774	14.115	19.013	22.873
NE	PSNR	30.857	23.902	21.667	21.584
	SSIM	0.881	0.706	0.798	0.610
	RMSE	5.541	12.345	15.549	20.601
ScSr1^[14]	PSNR	34.607	27.627	25.619	23.136
	SSIM	0.914	0.781	0.858	0.590
	RMSE	5.518	12.322	15.523	20.555
ODL1^[17]	PSNR	34.626	27.639	25.635	23.197
	SSIM	0.915	0.782	0.859	0.591
	RMSE	5.395	12.174	15.321	20.224
Ours1	PSNR	34.823	27.744	25.749	23.338
	SSIM	0.920	0.784	0.865	0.600
	RMSE	5.447	12.473	15.649	20.440
ScSr2^[14]	PSNR	34.709	27.711	25.535	23.242
	SSIM	0.917	0.781	0.858	0.593
	RMSE	5.459	12.190	15.621	20.418
ODL2^[17]	PSNR	34.720	27.732	25.579	23.255
	SSIM	0.916	0.782	0.859	0.594
	RMSE	5.395	12.104	15.501	20.224
Ours2	PSNR	34.823	27.793	25.648	23.338
	SSIM	0.920	0.784	0.863	0.600
	RMSE	5.577	12.202	15.700	20.663
ScS3^[14]	PSNR	34.589	27.729	25.539	23.168
	SSIM	0.912	0.782	0.856	0.588
	RMSE	5.540	12.186	15.689	20.615
ODL3^[17]	PSNR	34.592	27.735	25.541	23.171
	SSIM	0.913	0.783	0.857	0.589
	RMSE	5.513	12.172	15.520	20.576
Ours3	PSNR	34.636	27.745	25.636	23.187
	SSIM	0.915	0.784	0.861	0.591

Table 3. Numerical comparison of reconstruction for different images when the magnification factor is 3

Method	Evaluation index	Baby	Man	Butterfly	Baboon
	RMSE	7.601	15.334	23.192	22.934
Bicubic	PSNR	31.828	25.734	22.145	22.244
	SSIM	0.857	0.676	0.734	0.451
	RMSE	7.083	15.993	24.850	22.663
NE	PSNR	31.880	25.002	21.785	22.291
	SSIM	0.858	0.605	0.690	0.473
	RMSE	6.773	14.258	21.263	22.601
ScSr1^[14]	PSNR	32.864	26.447	22.905	22.378
	SSIM	0.876	0.711	0.754	0.490
	RMSE	6.751	14.203	21.218	22.512
ODL1^[17]	PSNR	32.872	26.450	22.919	22.407
	SSIM	0.877	0.712	0.755	0.491
	RMSE	6.735	14.109	21.040	22.497
Ours1	PSNR	32.893	26.462	22.992	22.413
	SSIM	0.878	0.713	0.757	0.492
	RMSE	6.747	14.488	21.091	22.552
ScSr2^[14]	PSNR	32.859	26.239	22.970	22.009
	SSIM	0.876	0.705	0.755	0.490
	RMSE	6.756	14.464	21.084	22.521
ODL2^[17]	PSNR	32.865	26.247	22.973	22.103
	SSIM	0.877	0.706	0.756	0.491
	RMSE	6.738	14.184	21.043	22.478
Ours2	PSNR	32.888	26.416	22.990	22.420
	SSIM	0.878	0.713	0.763	0.492
	RMSE	6.761	14.258	21.147	22.528
ScS3^[14]	PSNR	32.874	26.380	22.949	22.401
	SSIM	0.876	0.710	0.754	0.490
	RMSE	6.739	14.217	21.128	22.513
ODL3^[17]	PSNR	32.886	26.396	22.956	22.407
	SSIM	0.877	0.711	0.755	0.491
	RMSE	6.681	14.195	20.963	22.371
Ours3	PSNR	32.962	26.409	23.025	22.462
	SSIM	0.880	0.713	0.763	0.495

Table 4. Comparison of reconstruction for different images when the magnification factor is 4

Variance	Evaluationindex	Bicubic	NE	ScSr1	ODL1	Ours1	ScSr2	ODL2	Ours2	ScSr3	ODL3	Ours3
	RMSE	5.455	12.108	4.604	4.593	4.561	4.660	4.652	4.574	4.775	4.768	4.571
0	PSNR	34.713	28.669	36.231	36.239	36.299	36.119	36.127	36.276	35.907	35.912	36.280
	SSIM	0.911	0.788	0.925	0.926	0.927	0.924	0.925	0.926	0.923	0.924	0.926
	RMSE	6.942	7.001	5.903	5.889	5.727	6.386	6.377	6.244	6.597	6.584	6.261
0.0002	PSNR	32.622	30.757	34.057	34.068	34.309	33.368	33.372	33.556	33.088	33.092	33.532
	SSIM	0.908	0874	0.928	0.929	0.932	0.883	0.884	0.885	0.882	0.883	0.886
	RMSE	7.480	7.992	7.490	7.477	7.465	7.583	7.542	7.487	7.731	7.695	7.463
0.0005	PSNR	31.973	20.662	31.971	31.986	31.996	31.906	31.911	31.973	31.724	31.736	32.001
	SSIM	0.869	0.831	0.820	0.821	0.824	0.821	0.822	0.823	0.821	0.822	0.824

Table 5. Numerical comparison of reconstruction under different Gaussian noise environment

Density	Evaluationindex	Bicubic	NE	ScSr1	ODL1	Ours1	ScSr2	ODL2	Ours2	ScSr3	ODL3	Ours3
	RMSE	5.455	12.108	4.697	4.593	4.561	4.673	4.652	4.574	4.787	4.768	4.571
0	PSNR	34.713	28.669	36.231	36.239	36.299	36.114	36.127	36.276	35.906	35.912	36.280
	SSIM	0.911	0.788	0.925	0.926	0.927	0.924	0.925	0.926	0.923	0.924	0.926
	RMSE	6.969	7.145	5.902	5.888	5.821	5.895	5.889	5.727	6.501	6.486	6.084
0.0005	PSNR	32.588	32.077	34.057	34.066	34.165	34.059	34.068	34.309	33.219	33.228	33.783
	SSIM	0.906	0.873	0.927	0.928	0.929	0.928	0.929	0.932	0.922	0.923	0.929
	RMSE	7.305	7.661	6.872	6.844	6.799	7.101	7.066	6.964	7.165	7.159	6.862
0.001	PSNR	32.181	31.877	32.757	32.761	32.817	32.473	32.480	32.607	32.359	32.365	32.737
	SSIM	0.900	0.843	0.916	0.917	0.918	0.912	0.913	0.915	0.912	0.913	0.918

Table 6. Numerical comparison of reconstruction under different pepper & salt noise environment

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information