Hyperspectral Image Super-Resolution Network of Local-Global Attention Feature Reuse

Table 1. Influence of the number of module N on the network performance index
View table
Table 1. Influence of the number of module N on the network performance index
Index 6 7 8 9 10
PSNR↑ /dB 49.933 50.665 51.244 50.322 50.285
SSIM↑ 0.9954 0.9963 0.9964 0.9960 0.9960
SAM↓ 2.027 1.940 1.703 1.751 1.765
ERGA↓ 0.512 0.409 0.392 0.455 0.438

Table 2. Ablation study about the components

View table

Table 2. Ablation study about the components

Component	Different Combinations
FRM	×	√	√	√	√
LA	×	×	√	×	√
GC	×	×	×	√	√
PSNR↑ /dB	49.865	50.387	50.571	50.435	51.244
SSIM↑	0.9952	0.9956	0.9960	0.9960	0.9964
SAM↓	2.007	1.925	1.795	1.728	1.703
ERGA↓	0.595	0.547	0.485	0.477	0.392

Table 3. Comparison of algorithm results with different scale factors on CAVE dataset

View table

Table 3. Comparison of algorithm results with different scale factors on CAVE dataset

Scale	Metric	CSTF	UAL	TSFN	PZNet	Fusformer	DHIF	LGAR-Net
×4	PSNR /dB	42.300	43.964	48.853	50.288	49.106	50.169	51.244
	SSIM	0.9644	0.9920	0.9948	0.9950	0.9952	0.9952	0.9964
	SAM	7.716	3.042	1.869	1.982	1.997	1.833	1.703
	ERGA	2.093	1.629	0.461	0.530	0.521	0.470	0.392
×8	PSNR /dB	41.717	42.585	49.266	49.654	49.108	49.733	50.162
	SSIM	0.9665	0.9910	0.9943	0.9945	0.9944	0.9947	0.9947
	SAM	7.502	3.899	2.311	2.189	2.269	2.108	1.965
	ERGA	1.170	0.474	0.386	0.319	0.378	0.335	0.307
×16	PSNR /dB	39.824	40.105	46.533	46.994	46.322	46.865	47.128
	SSIM	0.9633	0.9886	0.9922	0.9920	0.9920	0.9922	0.9930
	SAM	7.329	4.257	2.921	2.881	2.961	2.848	2.768
	ERGA	0.892	0.381	0.311	0.287	0.324	0.265	0.246

Table 4. Comparison of algorithm results with different scale factors on Harvard dataset

View table

Table 4. Comparison of algorithm results with different scale factors on Harvard dataset

Scale	Metric	CSTF	UAL	TSFN	PZNet	Fusformer	DHIF	LGAR-Net
×4	PSNR /dB	42.362	44.853	48.993	49.382	48.856	49.263	49.962
	SSIM	0.9722	0.9865	0.9874	0.9873	0.9873	0.9873	0.9882
	SAM	8.257	4.762	3.488	3.250	3.422	3.488	3.085
	ERGA	5.693	1.855	1.732	1.640	1.662	1.539	1.218
×8	PSNR /dB	41.233	43.368	48.661	48.864	48.203	48.835	49.210
	SSIM	0.9706	0.9837	0.9850	0.9849	0.9850	0.9850	0.9862
	SAM	9.214	5.011	3.763	3.901	3.955	3.858	3.317
	ERGA	3.847	1.025	0.839	0.807	0.822	0.796	0.752
×16	PSNR /dB	39.822	42.563	47.299	47.262	46.958	47.235	47.386
	SSIM	0.9521	0.9788	0.9803	0.9801	0.9803	0.9803	0.9812
	SAM	9.621	5.812	4.344	4.219	4.425	4.268	4.105
	ERGA	2.102	0.842	0.662	0.678	0.666	0.655	0.602

Tools

Get Citation

Copy Citation Text

Size Wang, Xin Guan, Qiang Li. Hyperspectral Image Super-Resolution Network of Local-Global Attention Feature Reuse[J]. Acta Optica Sinica, 2023, 43(21): 2115001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 2, 2023

Accepted: May. 31, 2023

Published Online: Nov. 8, 2023

The Author Email: Li Qiang (liaiqng@tju.edu.cn)

DOI:10.3788/AOS230613

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Influence of the number of module N on the network performance index

Table 1. Influence of the number of module N on the network performance index

Table 2. Ablation study about the components

Table 2. Ablation study about the components

Table 3. Comparison of algorithm results with different scale factors on CAVE dataset

Table 3. Comparison of algorithm results with different scale factors on CAVE dataset

Table 4. Comparison of algorithm results with different scale factors on Harvard dataset

Table 4. Comparison of algorithm results with different scale factors on Harvard dataset