Fig. 1.

Strategy to evaluate algorithms used to denoise a noisy tomogram Y containing n 2D slices (A), i.e., $\mathit{Y}={\left(Y_i\right)}_{i=1}^n$, where the index i represents the i^th 2D slice in the noisy tomogram. Ŷ is the subset of Y, which is selected from a comparable region of each tomographic slice that contains no cells and hence no biological data (black box, enlarged on right). N represents the tomographic noise (N = Y − X, where X denotes the denoised tomogram). N̂ is the subset of N similar to Ŷ, where no biological data exists. A Gaussian fit computed with the shown mean and variance to the noise samples in Ŷ is shown on lower right (representative slice). Scale bar: 100 nm (B) Iterative process to identify the best denoising algorithm with optimized parameters, for a particular tomogram. Quantitative assessment was performed using analysis such as the KL-distance based GOF test, Fourier Ring Correlation and Single-Image SNR (see Methods for details).