Extended Data Fig. 4: Simulation comparisons of different SOFI methods and SACD under different SNRs using ring structures.
From: Enhanced detection of fluorescence fluctuations for high-throughput super-resolution imaging
(a) The synthetic ring structures were convolved with a 110 nm PSF and down-sampled 3 times (pixel size 30 nm) as ground truth (GT, Methods). (b-f) Wide-field (WF, b) raw images and SOFI-wavelet (c), SOFI-RL (d), RD-Covar (e), and SACD (f) reconstructions using 20-frame under ‘High’ (right), ‘Medium’ (middle), and ‘Low’ (left) SNR conditions. To generate the raw WF image stacks, the structures were convolved with a 220 nm PSF, and down-sampled 3 times before adding mixture noise under three different levels, in which the mixture noise includes cytosol background, out-of-focus light, Poisson noise, Gaussian readout noise, and baseline background. Yellow numbers represent the minimum resolvable diameters of ring structures according to the RQR criterion. The sub-images at the upper right corner are the enlarged view from white box. The sub-images at the lower right corner are the intensity profiles and multiple Gaussian fitting from white line. (g) The SSIM curves of different methods under ‘High’ (right), ‘Medium’ (middle), and ‘Low’ (left) SNR conditions. (h) The calculation method of ring quality ratio (RQR) criterion. (i) The STD curves of different methods under SNR conditions. Details in Supplementary Note 2. Experiments were repeated ten times independently with similar results. Scale bars: (a) 300 nm; (a, inset) 150 nm.
