Autofocus method based on frequency domain cross-correlation in a structured illumination microscopy

Precise focusing is a challenge for live-cell super-resolution structured illumination microscopy (SR-SIM), especially in the case of low signal-to-noise ratio of raw images. Here, we propose an integrated autofocus algorithm (IT-SICC) that combines frequency domain cross-correlation with an Improved Tenengrad algorithm to achieve high-precision autofocus. Due to leveraging the prior characteristics of the SIM imaging fringe pattern, the IT-SICC outperforms existing image-based autofocus methods in terms of focusing accuracy, noise robustness, and independence from morphological features. The IT-SICC method achieves a focusing resolution with a 50 nm step size in fluorescent microsphere experiments, nearly four times more precise than using the IT autofocus method. SR-SIM imaging with the IT-SICC method for living MCF7 cells with varying structural features shows improved SIM image reconstruction quality with the increased PSNR from 32.1 to 35.5 and SSIM from 0.82 to 0.93, demonstrating that the IT-SICC method reduces the defocus blur-induced reconstruction artifacts.