Fig. 4: Systematic deblurring analysis of F-actin and ER.

a Representative deblurred (DB) images derived from blurry F-actin. Subsequent to each blurry image, seven distinct Deblurred (DB) renditions, realized through varied settings, are exhibited. Beneath each DB image, the corresponding difference image is presented, strategically aligned against the ground truth (GT) for direct comparison. The Mean Absolute Error (MAE) for each difference image was meticulously quantified and annotated at the upper left corner, facilitating a precise evaluation of the deblurring efficacy. Scale bar, 4 μm. b Comparison of DB images of ERs inferred by Src and ours. Blur and GT images are shown for reference. Lower row shows the magnified images of the boxed regions in the upper images. Scale bar 1 μm. c, d Statistical comparison of the traditional and HitchLearning in terms of PSNR, SSIM, MSE and MAE for F-actin (c) and ER (d). Combining visual results with a comprehensive quantitative analysis across seven distinct settings, our novel paradigm, HitchLearning, achieves unparalleled performance in single-image deblurring specifically for bioimaging applications.