Fig 3. Characterization of the CNN-predicted image quality improvement.
Input and predicted images by 64- and 16-channel sparse array, and the ground truth images by the original 128-channel dense array of two representative teeth: (a) 2nd pre-molar and (b) 3rd molar. (c) SSIM, (d) MSE, and (e) PSNR of the 128-channel ground truth image versus 16- and 64-channel input and predicted images of all 100 pairs of images in the entire test dataset. ****: p < 0.0001. SSIM: structural similarity index measure. MSE: mean squared error. PSNR: peak signal-to-noise ratio.