Table 1 Summary of the performance measure (SSIM) of the proposed deep learning (DL) approach, as obtained through a 5-fold cross validation procedure repeated 100 times.

From: Deep learning approach for denoising low-SNR correlation plenoptic images

Sampling ratio (S)

SSIM label-input

SSIM label-output

0.025%

0.050 ± 0.001

0.630 ± 0.150

0.25%

0.090 ± 0.010

0.680 ± 0.140

5%

0.280 ± 0.010

0.920 ± 0.070

  1. The SSIM values obtained for the 3 considered sampling ratios are reported with the corresponding standard deviations. Input refers to images obtained by CPI; output to images obtained after implementing our DL model.
Back to article page