Chinese Journal of Liquid Crystals and Displays, Volume. 39, Issue 9, 1214(2024)
Lens-free adversarial encoded imaging method based on convolutional attention mechanism
Fig. 1. （a）Coded aperture imaging setup；（b）CAM-GAN framework；（c）Generative network in figure（b）；（d）Discriminative network in figure（b）.
Fig. 2. Reconstructed image comparison of different algorithms（simulation verification）.（a）OC=0.15 m，CI=0.1 m；（b）OC=0.22 m，CI=0.1 m.
Fig. 3. Settings of simulation experiments.（a）Details of aperture coding mask experiments；（b）Photograph of setup，The corresponding input image from the senor is shown in figure（a）.
Fig. 4. Reconstructed image comparison of different algorithms（experiment verification），OC=45 mm，CI=8 mm.
Table 1. Comparison of evaluation indexes of different algorithms
View tableView in ArticleTable 1. Comparison of evaluation indexes of different algorithms
Method PSNR/dB SSIM MAE Deconvolution 16.591 0 0.560 5 0.121 1 Richardson-Lucy 16.146 7 0.583 4 0.132 1 UNet 19.564 6 0.708 3 0.083 4 UNet-GAN 20.799 9 0.709 4 0.069 1 CAM-UNet 19.632 5 0.710 8 0.083 4 CAM-GAN 21.118 5 0.712 7 0.066 2
Table 2. Comparison of evaluation indexes of CAM-GAN algorithm under different distance
View tableView in ArticleTable 2. Comparison of evaluation indexes of CAM-GAN algorithm under different distance
Distance PSNR/dB SSIM MAE OC=0.15 m，CI=0.1 m 21.118 5 0.712 7 0.066 2 OC=0.22 m，CI=0.1 m 20.692 3 0.718 9 0.069 0
Table 3. Comparison of evaluation indexes of different algorithms
View tableView in ArticleTable 3. Comparison of evaluation indexes of different algorithms
Method PSNR/dB SSIM MAE UNet 31.456 4 0.987 9 0.018 0 UNet-GAN 32.446 2 0.990 5 0.014 6 CAM-UNet 31.724 3 0.985 1 0.019 4 CAM-GAN 33.254 7 0.990 9 0.013 5
Get Citation
Copy Citation Text
Hailun WANG, Yueke WANG. Lens-free adversarial encoded imaging method based on convolutional attention mechanism[J]. Chinese Journal of Liquid Crystals and Displays, 2024, 39(9): 1214
Category:
Received: Sep. 26, 2023
Accepted: --
Published Online: Nov. 13, 2024
The Author Email: Yueke WANG (ykwang@jiangnan.edu.cn)
