Fig. 5: Systematic super-resolution analysis for F-actin and ER.

a, b Representative SR images reconstructed by Src (a) and ours (b) from raw images from ER and F-actin. c The corresponding GT images. Scale bar, 1 μm for ER and F-actin. d Comparison of SR images of ERs inferred by Src and ours. LR and GT images are shown for reference. right row shows the magnified images of the boxed regions in the left images. Scale bar 5 μm; 1 μm for magnified images. e Intensity profiles of F-actin (2× upscaling) along the lines indicated by the two arrowheads in the images of (a) Src (orange), (b) ours (green) and (c) GT (white). The arrows indicate that the structures reconstructed by ours are closer to GT than those reconstructed by Src. f Representative SR images of F-actin generated by HitchLearning. Bottom: a fraction of the corresponding LR image. Scale bar, 2 μm. Through a combination of visual assessments and quantitative analysis across seven distinct settings, our innovative paradigm, HitchLearning, was demonstrated to achieve the most effective performance in Single-Image Super-Resolution (SISR) specifically for live-cell imaging. For comparison purposes, the low-resolution (LR) images were resized to match the dimensions of the super-resolution (SR) images.