Fig. 2

Comparative simulations of different methods for parameter estimation and super-resolution reconstruction. a The wave vector errors of PCA-SIM and COR under different noise environments. b The initial phase errors of different methods (PCA-SIM, COR, POP, ACR and IRT) under different noise environments. c The wide-field images of the samples under different noise conditions and the super-resolution image reconstructed by PCA-SIM. Images in the middle circular area are the wide-field images when the noise power is 25dBW and 45dBW respectively, and the image in the surrounding rectangular area is the super-resolution image acquired by PCA-SIM when the noise is 25dBW. The magnified label image corresponding to the blue boxed region is shown in the top left. d Magnified wide-field images (left) and super-resolution images from the boxed regions in c obtained by different methods when the noise is 25dBW (‘None’ means no parameter estimation). To quantify the reconstruction quality, we calculate the structural similarity index measure (SSIM) between the reconstruction results and Label, which are 0.8377, 0.8383, 0.8468, 0.9736 and 0.9739 for POP, ACR, IRT, COR, and PCA, respectively. e Magnified wide-field images (left) and super-resolution images from the boxed regions in c obtained by different methods when the noise is 35dBW. The SSIM are 0.7924, 0.7919, 0.8052, 0.9152 and 0.9661 for POP, ACR, IRT, COR, and PCA, respectively. f Magnified wide-field images (left) and super-resolution images from the boxed regions in c obtained by different methods when the noise is 45dBW. The SSIM are 0.6872, 0.6785, 0.6850, 0.7177 and 0.9451 for POP, ACR, IRT, COR, and PCA, respectively. g Intensity profiles along the light blue line in d-f (normalized to maximum). Simulations were repeated ten times independently with similar results. Scale bars: 1 \(\mu\)m (c)-(f)