Acta Photonica Sinica, Volume. 52, Issue 4, 0430002(2023)
Hyperspectral Image Denoising Based on Fast Tri-factorization and Group Sparsity Regularized
Figures & Tables(19)
Fig. 3. Denoising results on band 65 of Indian Pines dataset in case 5
Fig. 4. Denoising results on band 66 of Washington DC dataset in case 5
Fig. 5. Comparison diagram of PSNR and SSIM values of six models in different bands on the Indian Pines dataset under different noise conditions
Fig. 6. Comparison diagram of PSNR and SSIM values of six models in different bands on the Washington DC dataset under different noise conditions
Fig. 8. Spectrum of Washington DC dataset at pixel（10，180）in case 6
Fig. 9. Denoising results of the 150th band of real data by using different models
Fig. 10. Denoising results of the 162th band of real data by using different models
Table 1. Different noise scenarios
View tableView in ArticleTable 1. Different noise scenarios
Noise
case
Gaussian noise Gaussian noise Impulse noise Deadline Stripe Deadline Stripe （mean 0，variance 0.1） （mean 0，variance 0~0.2 random） （percentage 0~0.2 random） （number 3~10 random） （number 3~10 random） （number 3~10 random） （number 3~10 random） All bands All bands All bands 40% bands 40% bands 20% bands 20% bands Case1 √ - - - - - - Case2 - √ - - - - - Case3 - √ √ - - - - Case4 - √ √ √ - - - Case5 - √ - - √ - - Case6 - √ √ - - √ √
Table 2. Quantitative evaluation results of simulated experiment
View tableView in ArticleTable 2. Quantitative evaluation results of simulated experiment
Dataset Noisecase Level Index noisy LRMR LRTDTV LLRSSTV LRTDGS LLxRGTV FTFGS Indian Pines Case1 G=0.1 MPSNR 20.258 8 36.689 0 40.480 0 38.473 7 42.837 0 39.959 6 45.732 6 MSSIM 0.400 4 0.947 0 0.990 0 0.953 7 0.996 0 0.991 6 0.997 2 MFSIM 0.489 2 0.945 0 0.986 0 0.951 4 0.994 6 0.987 3 0.997 8 Case2 G MPSNR 22.805 2 37.196 3 42.385 0 40.274 5 43.282 7 41.727 6 46.684 7 MSSIM 0.478 1 0.951 1 0.993 0 0.960 1 0.996 8 0.995 0 0.996 9 MFSIM 0.557 4 0.950 9 0.991 0 0.966 3 0.996 8 0.993 6 0.997 8 Case3 G+P MPSNR 14.623 4 35.889 0 40.978 0 39.726 3 42.325 8 40.466 5 45.501 6 MSSIM 0.261 7 0.938 0 0.991 0 0.961 2 0.996 1 0.993 7 0.996 9 MFSIM 0.435 0 0.938 0 0.987 0 0.962 7 0.995 7 0.991 1 0.997 9 Case4 G+P+Dealine MPSNR 14.133 4 35.102 0 40.324 0 35.504 2 40.245 6 38.969 7 42.242 7 MSSIM 0.253 6 0.940 0 0.989 0 0.958 1 0.993 0 0.991 3 0.993 5 MFSIM 0.426 9 0.940 0 0.986 0 0.954 5 0.992 2 0.988 3 0.995 0 Case5 G+Stripe MPSNR 22.009 8 37.112 2 42.202 3 39.940 1 43.064 8 41.197 6 46.859 0 MSSIM 0.469 4 0.950 5 0.992 2 0.953 6 0.996 6 0.994 2 0.997 5 MFSIM 0.551 4 0.951 1 0.989 4 0.960 2 0.996 6 0.992 7 0.998 3 Case6 G+P+Dealine+Stripe MPSNR 14.343 1 35.488 1 40.559 4 37.132 4 41.767 2 39.604 43.524 5 MSSIM 0.256 4 0.940 8 0.990 0 0.947 7 0.995 5 0.992 4 0.993 6 MFSIM 0.429 5 0.941 1 0.986 4 0.950 6 0.994 5 0.989 3 0.995 3 Washington
DC
Case1 G=0.1 MPSNR 20.851 5 34.292 0 34.450 8 36.776 2 36.513 1 35.179 0 37.385 0 MSSIM 0.365 9 0.912 3 0.924 9 0.948 0 0.948 5 0.934 3 0.953 5 MFSIM 0.692 9 0.956 6 0.952 6 0.969 8 0.969 3 0.958 2 0.972 3 Case2 G MPSNR 23.522 7 35.013 2 35.937 8 38.553 4 37.772 4 36.633 2 38.637 8 MSSIM 0.465 2 0.914 4 0.943 5 0.961 3 0.962 4 0.951 2 0.967 2 MFSIM 0.729 5 0.957 7 0.966 1 0.976 9 0.977 6 0.967 7 0.979 1 Case3 G+P MPSNR 14.249 7 33.865 8 35.277 6 36.242 9 37.123 7 35.752 8 37.885 3 MSSIM 0.194 4 0.896 0 0.942 3 0.951 3 0.957 5 0.942 9 0.961 1 MFSIM 0.564 6 0.949 1 0.965 1 0.968 3 0.975 3 0.963 0 0.975 4 Case4 G+P+Dealine MPSNR 14.181 0 33.428 1 35.432 1 36.334 4 36.156 0 35.678 3 37.017 6 MSSIM 0.192 6 0.888 3 0.945 0 0.951 5 0.951 7 0.943 3 0.958 4 MFSIM 0.562 6 0.944 2 0.967 0 0.969 8 0.971 5 0.964 5 0.971 6 Case5 G+Stripe MPSNR 23.101 4 34.858 6 35.889 0 37.837 1 37.714 5 36.501 8 38.551 3 MSSIM 0.458 6 0.913 4 0.949 5 0.943 1 0.961 8 0.950 1 0.966 2 MFSIM 0.724 5 0.957 0 0.969 4 0.966 5 0.977 3 0.966 9 0.978 8 Case6 G+P+Dealine+Stripe MPSNR 14.227 0 33.706 5 35.217 6 36.251 2 36.222 6 35.861 7 37.413 2 MSSIM 0.192 9 0.894 7 0.941 7 0.930 4 0.952 5 0.945 2 0.960 2 MFSIM 0.562 9 0.947 7 0.964 6 0.958 6 0.971 8 0.965 4 0.973 7
Table 3. Running time in simulation experiment
View tableView in ArticleTable 3. Running time in simulation experiment
Dataset Noise Case Running time/s Improvement/% LRMR LRTDTV LRTDGS LLxRGTV LLRSSTV FTFGS Indian Pines Case1 83.11 85.88 77.54 198.68 186.24 80.46 56.80 Case2 84.19 85.60 79.14 242.13 215.98 81.03 62.48 Case3 84.81 93.67 83.10 199.84 220.28 80.94 63.26 Case4 84.32 94.04 80.67 245.23 185.66 81.39 56.16 Case5 103.16 85.50 78.89 237.08 247.98 80.84 67.40 Case6 84.00 87.29 79.95 199.91 169.45 81.58 51.86 Washington DC Case1 245.01 217.27 176.39 609.38 464.68 226.09 51.34 Case2 233.31 218.65 179.92 606.24 395.82 226.74 42.72 Case3 174.51 218.49 179.86 605.45 407.70 228.19 44.03 Case4 174.46 220.95 182.89 599.88 376.45 235.85 37.35 Case5 260.70 223.36 183.40 601.73 364.53 227.34 37.63 Case6 254.41 221.96 177.19 610.79 403.63 229.21 43.21
Tools
Get Citation
Copy Citation Text
Xiaoyu GAO, Jingyuan BAI, Yangzhi HUANG, Jifeng NING. Hyperspectral Image Denoising Based on Fast Tri-factorization and Group Sparsity Regularized[J]. Acta Photonica Sinica, 2023, 52(4): 0430002
Paper Information
Category:
Received: Oct. 9, 2022
Accepted: Jan. 3, 2023
Published Online: Jun. 21, 2023
The Author Email: NING Jifeng (njf@nwafu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20