Multi-scale dense feature fusion network for image super-resolution

Deqiang CHENG; Jiamin ZHAO; Qiqi KOU; Liangliang CHEN; Chenggong HAN

doi:10.37188/OPE.20223020.2489

Journals >Optics and Precision Engineering >Volume 30 >Issue 20 >Page 2489 > Article

Optics and Precision Engineering
Vol. 30, Issue 20, 2489 (2022)

Multi-scale dense feature fusion network for image super-resolution

Deqiang CHENG¹, Jiamin ZHAO¹, Qiqi KOU^2,*, Liangliang CHEN¹, and Chenggong HAN¹

Author Affiliations

¹School of Information and Control Engineering， China University of Mining and Technology， Xuzhou226， China

²School of Computer Science and Technology， China University of Mining and Technology， Xuzhou1116， China

show less

DOI: 10.37188/OPE.20223020.2489 Cite this Article

Deqiang CHENG, Jiamin ZHAO, Qiqi KOU, Liangliang CHEN, Chenggong HAN. Multi-scale dense feature fusion network for image super-resolution[J]. Optics and Precision Engineering, 2022, 30(20): 2489 Copy Citation Text

show less

Fig. 1. Multi-scale dense feature fusion network

Download full size | View in the Article

Fig. 2. Multi-scale feature fusion residual block

Download full size | View in the Article

Fig. 3. Upsampling network

Download full size | View in the Article

Fig. 4. Reconstruction results of Img040 in Urban100

Download full size | View in the Article

Fig. 5. Reconstruction results of Img056 in Urban100

Download full size | View in the Article

Fig. 6. Reconstruction results of Img092 in Urban100

Download full size | View in the Article

Fig. 7. PSNR results of different D, C and G models

Download full size | View in the Article

Fig. 8. Reconstruction results of Img056 in Urban100

Download full size | View in the Article

Fig. 9. Reconstruction results of Img081 in Urban100

Download full size | View in the Article

Fig. 10. PSNR and parameters on Set5 （×4） dataset of different models

Download full size | View in the Article

Model	层次特征融合	密集特征融合	PSNR/dB	SSIM
NSFB	√	×	32.30	0.929 9
	×	√	32.35	0.930 5
Resblock	√	×	31.50	0.922 0
	×	√	31.53	0.922 2
RDN	√	×	32.22	0.928 9
	×	√	32.29	0.929 6
MSRB	√	×	32.29	0.929 7
	×	√	32.36	0.930 4
MFRB	√	×	32.35	0.930 3
	×	√	32.41	0.932 2

Table 1. Effects of different connection modes and different feature extraction modules on model performance

View in the Article

模型	缩放因子	Set5		Set14		BSD100		Urban100
模型	缩放因子	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM
Bicubic	×2	33.66	0.929 9	30.24	0.868 7	29.56	0.843 1	26.88	0.840 1
SRCNN	×2	36.66	0.954 2	32.43	0.906 3	31.36	0.887 9	29.50	0.894 6
VDSR	×2	37.54	0.958 7	33.03	0.912 4	31.90	0.896 0	30.76	0.914 0
DRCN	×2	37.63	0.958 4	33.06	0.910 8	31.85	0.894 7	30.76	0.914 7
LapSRN	×2	37.53	0.959 1	33.08	0.910 9	31.80	0.894 9	30.41	0.911 2
MSRN	×2	38.08	0.960 5	33.74	0.917 0	32.23	0.901 3	32.22	0.932 6
IMDN	×2	38.00	0.960 5	33.63	0.917 7	32.19	0.899 6	32.17	0.928 3
OISR-SK2	×2	38.12	0.960 9	33.80	0.919 3	32.26	0.900 6	32.48	0.931 7
LatticeNet	×2	38.15	0.961 0	33.78	0.919 3	32.25	0.900 5	32.43	0.930 2
DID-D5	×2	38.15	0.961 0	33.77	0.919 0	32.27	0.900 6	32.38	0.930 5
MDFN	×2	38.14	0.961 0	33.83	0.919 6	32.27	0.900 6	32.41	0.931 0
Bicubic	×3	30.39	0.868 2	27.54	0.773 6	27.21	0.738 4	24.46	0.734 4
SRCNN	×3	32.75	0.909 0	29.30	0.821 5	28.41	0.786 3	26.24	0.798 9
VDSR	×3	33.66	0.921 3	29.77	0.831 4	28.82	0.797 6	27.14	0.827 9
DRCN	×3	33.85	0.921 5	29.89	0.831 7	28.81	0.795 4	27.16	0.831 1
LapSRN	×3	33.82	0.922 7	29.89	0.832 0	28.83	0.797 3	27.08	0.827 2
MSRN	×3	34.38	0.926 2	30.34	0.839 5	29.08	0.804 1	28.08	0.855 4

Table 2. Index comparison under benchmark dataset when scaling factor is 2， 3 and 4

View in the Article

模型	缩放因子	Set5		Set14		BSD100		Urban100
模型	缩放因子	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM
IMDN	×3	34.36	0.927 0	30.32	0.841 7	29.09	0.804 6	28.17	0.851 9
OISR-SK2	×3	34.55	0.928 2	30.46	0.844 3	29.18	0.807 5	28.50	0.859 7
LatticeNet	×3	34.53	0.928 1	30.39	0.842 4	29.15	0.805 9	28.33	0.853 8
DID-D5	×3	34.55	0.928 0	30.49	0.844 6	29.19	0.806 9	28.39	0.856 6
MDFN	×3	34.60	0.928 4	30.50	0.844 9	29.21	0.807 5	28.52	0.859 1
Bicubic	×4	28.42	0.810 4	26.00	0.701 9	25.96	0.667 4	23.14	0.657 0
SRCNN	×4	30.48	0.862 8	27.49	0.750 3	26.90	0.710 1	24.53	0.722 1
VDSR	×4	31.35	0.883 0	28.01	0.768 0	27.29	0.725 1	25.18	0.754 3
DRCN	×4	31.56	0.881 0	28.15	0.762 0	27.24	0.715 0	25.15	0.753 0
LapSRN	×4	31.54	0.885 5	28.19	0.772 0	27.32	0.728 0	25.21	0.755 3
MSRN	×4	32.07	0.890 3	28.60	0.775 1	27.52	0.727 3	26.04	0.789 6
IMDN	×4	32.21	0.894 8	28.58	0.781 1	27.56	0.735 3	26.04	0.783 8
OISR-SK2	×4	32.32	0.896 5	28.72	0.784 3	27.66	0.739 0	26.37	0.795 3
LatticeNet	×4	32.30	0.896 2	28.68	0.783 0	27.62	0.736 7	26.25	0.787 3
DID-D5	×4	32.33	0.896 8	28.75	0.785 2	27.68	0.738 6	26.36	0.793 3
MDFN	×4	32.41	0.897 6	28.78	0.786 0	27.69	0.739 3	26.39	0.794 4

Table 2. Index comparison under benchmark dataset when scaling factor is 2， 3 and 4

View in the Article

模型	参数量/M	Flops/G	PSNR/dB	SSIM
MSRN	6.08	107.27	32.07	0.890 3
OISR-SK2	5.51	117.41	32.32	0.896 5
DID-D5	5.21	93.03	32.33	0.896 8
MDFN	4.89	87.76	32.41	0.897 6

Table 3. Complexity and performance of different models

Deqiang CHENG, Jiamin ZHAO, Qiqi KOU, Liangliang CHEN, Chenggong HAN. Multi-scale dense feature fusion network for image super-resolution[J]. Optics and Precision Engineering, 2022, 30(20): 2489

Download Citation

Tools

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享