Lightweight Super-Resolution Image-Reconstruction Model with Adaptive Residual Attention

Ming Jiang; Qingsheng Xiao; Jianbing Yi; Feng Cao

doi:10.3788/LOP202259.1610007

Laser & Optoelectronics Progress, Volume. 59, Issue 16, 1610007(2022)

Lightweight Super-Resolution Image-Reconstruction Model with Adaptive Residual Attention

Ming Jiang, Qingsheng Xiao, Jianbing Yi^*, and Feng Cao

Author Affiliations

School of Information Engineering, Jiangxi University of Science and Technology, Ganzhou 341000, Jiangxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Adaptive residual attention super-resolution network structure

Download full size

Fig. 2. Coordinate attention

Download full size

Fig. 3. Heat maps of attention map under different network layers

Download full size

Fig. 4. Adaptive residual attention information extraction module

Download full size

Fig. 5. Effect of number of ARB modules on model performance

Download full size

Fig. 6. Correspondence between number of model parameters and PSNR values for each algorithm

Download full size

Fig. 7. Results of different algorithms for the “253027.png” in BSD100 dataset under 4× magnification

Download full size

Fig. 8. Results of different algorithms for the “img_046” in Urban100 dataset under 4× magnification

Download full size

Fig. 9. Results of different algorithms for the “img_012” in Urban100 dataset under 4× magnification

Download full size

Table 1. Parameter setting of the proposed network structure

View table

Table 1. Parameter setting of the proposed network structure

Layer_name	Kernel_size	Input_channel	Output_channel
Conv1	（3，3）	3	64
［ARB］×20（RB）	（1，1）	3	32
	（3，3）	32	32
	（3，3）	32	32
［ARB］×20（CoordA）	（1，1）	64	32
	（1，1）	32	16
	（1，1）	16	32
	（1，1）	16	32
	（1，1）	64	64
Upconv1	（3，3）	64	256
Upconv2	（3，3）	64	256
Conv_out	（3，3）	64	3

Table 2. Number of parameters of different module combinations and PSNR values under Set5 dataset at a magnification of 4
View table
Table 2. Number of parameters of different module combinations and PSNR values under Set5 dataset at a magnification of 4
Module Number of parameters /10³ PSNR /dB
Baseline 1517 25.98
Baseline+CA 1526 26.02
Baseline+CoordA 1544 26.03
ARASR（ours） 922 26.09

Table 3. Average PSNR and SSIM of different SR algorithms under 2×, 3×, and 4× magnification on the four datasets

View table

Table 3. Average PSNR and SSIM of different SR algorithms under 2×, 3×, and 4× magnification on the four datasets

Method	Scale	Number of parameters /10³	Set5 PSNR /dB SSIM	Set14 PSNR /dB SSIM	BSD100 PSNR /dB SSIM	Urban100 PSNR /dB SSIM
Bicubic	2×		33.66 0.9299	30.24 0.8688	29.56 0.8431	26.88 0.8403
SRCNN		57	36.66 0.9542	32.45 0.9067	31.36 0.8879	29.50 0.8946
FSRCNN		13	37.00 0.9558	32.63 0.9088	31.53 0.8920	29.88 0.9020
VDSR		666	37.53 0.9587	33.03 0.9124	31.90 0.8942	30.75 0.9133
DRCN		1774	37.63 0.9588	33.04 0.9118	31.85 0.8942	30.75 0.9133
LapSRN		251	37.52 0.9591	32.99 0.9124	31.80 0.8952	30.41 0.9103
DRRN		298	37.74 0.9591	33.23 0.9136	32.05 0.8973	31.23 0.9188
MemNet		678	37.78 0.9597	33.28 0.9142	32.08 0.8978	31.31 0.9195
EDSR-baseline		1370	37.99 0.9604	33.57 0.9175	32.16 0.8994	31.98 0.9272
SRMDNF		1511	37.79 0.9601	33.32 0.9159	32.05 0.8985	31.33 0.8204
CARN		1592	37.76 0.9590	33.52 0.9166	32.09 0.8978	31.92 0.9256
ARASR-s		307	37.88 0.9602	33.49 0.9171	32.12 0.8991	31.92 0.9266
ARASR		774	38.02 0.9606	33.71 0.9189	32.19 0.9000	32.24 0.9297
Bicubic	3×		30.39 0.8682	27.55 0.7742	27.21 0.7385	24.46 0.7349
SRCNN		57	32.75 0.9090	29.30 0.8215	28.41 0.7863	26.24 0.7989
FSRCNN		13	33.18 0.9140	29.37 0.8240	28.53 0.7910	26.43 0.8080
VDSR		666	33.66 0.9213	29.77 0.8314	28.82 0.7976	27.14 0.8279
DRCN		1774	33.82 0.9226	29.76 0.8311	28.80 0.7963	27.15 0.8276
LapSRN		502	33.81 0.9220	29.79 0.8325	28.82 0.7980	27.07 0.8275
DRRN		298	34.03 0.9244	29.96 0.8349	28.96 0.8004	27.53 0.8378
MemNet		678	34.09 0.9248	30.00 0.8350	28.96 0.8001	27.56 0.8376
EDSR-baseline		1555	34.37 0.9270	30.28 0.8417	29.09 0.8052	28.15 0.8527
SRMDNF		1528	34.12 0.9254	30.04 0.8382	28.97 0.8025	27.57 0.8398
CARN		1592	34.29 0.9255	30.29 0.8407	29.06 0.8034	28.06 0.8493
ARASR-s		379	34.24 0.9262	30.27 08411	29.03 0.8038	27.98 0.8488
ARASR		959	34.37 0.9270	30.39 0.8433	29.11 0.8056	28.29 0.8552
Bicubic	4×		28.42 0.8104	26.00 0.7027	25.96 0.6675	23.14 0.6577
SRCNN		57	30.48 0.8628	27.50 0.7513	26.90 0.7101	24.52 0.7221
FSRCNN		13	30.72 0.8660	27.61 0.7550	26.98 0.7150	24.62 0.7280
VDSR		665	31.35 0.8838	28.01 0.7674	27.29 0.7251	25.18 0.7524
DRCN		1774	31.53 0.8854	28.02 0.7670	27.23 0.7233	25.14 0.7510
LapSRN		813	31.54 0.8852	28.09 0.7700	27.32 0.7275	25.21 0.7562
DRRN		297	31.68 0.8888	28.21 0.7720	27.38 0.7284	25.44 0.7638
MemNet		678	31.74 0.8893	28.26 0.7723	27.40 0.7281	25.50 0.7630
EDSR-baseline		1518	32.09 0.8938	28.58 0.7813	27.57 0.7357	26.04 0.7849
SRMDNF		1552	31.96 0.8925	28.35 0.7787	27.49 0.7337	25.68 0.7731
CARN		1592	32.13 0.8937	28.60 0.7806	27.58 0.7349	26.07 0.7837
ARASR-s		365	32.06 0.8932	28.50 0.7790	27.51 0.7342	25.93 0.7807
ARASR		922	32.20 0.8951	28.61 0.7821	27.58 0.7367	26.15 0.7876

Tools

Get Citation

Copy Citation Text

Ming Jiang, Qingsheng Xiao, Jianbing Yi, Feng Cao. Lightweight Super-Resolution Image-Reconstruction Model with Adaptive Residual Attention[J]. Laser & Optoelectronics Progress, 2022, 59(16): 1610007

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Jun. 1, 2021

Accepted: Jun. 27, 2021

Published Online: Jul. 22, 2022

The Author Email: Jianbing Yi (yijianbing8@163.com)

DOI:10.3788/LOP202259.1610007

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Parameter setting of the proposed network structure

Table 1. Parameter setting of the proposed network structure

Table 2. Number of parameters of different module combinations and PSNR values under Set5 dataset at a magnification of 4

Table 2. Number of parameters of different module combinations and PSNR values under Set5 dataset at a magnification of 4

Table 3. Average PSNR and SSIM of different SR algorithms under 2×, 3×, and 4× magnification on the four datasets

Table 3. Average PSNR and SSIM of different SR algorithms under 2×, 3×, and 4× magnification on the four datasets