Optical-parameter-conditioned one-shot autofocus for multi-objective-lens microscopy

Autofocus is essential for ensuring image quality and scanning efficiency in microscopy. Although learning-based one-shot autofocus methods can estimate the defocus distance from a single defocused image and have shown encouraging progress, most are developed and validated under a fixed-objective-lens setting, making them difficult to adapt directly to practical multi-objective-lens microscopy. In particular, a model trained for one objective lens can incur substantially larger prediction errors when applied to another. In this paper, we propose an optical-parameter-conditioned one-shot autofocus method, termed OPC-OSAF. By introducing an optical-parameter-conditioned recalibration module (OPCRM), OPC-OSAF incorporates magnification, numerical aperture, and immersion-medium refractive index into intermediate feature recalibration, allowing a single model to condition its predictions on the objective-lens configuration. We construct MOL-DS, a multi-objective-lens defocus-stack dataset covering seven objective lenses and approximately 60,000 labeled defocused microscopy images. To the best of our knowledge, this work provides one of the first systematic evaluations of one-shot autofocus in multi-objective-lens microscopy. Experimental results show that OPC-OSAF outperforms several representative baselines in terms of depth-of-field (DoF)-normalized error, prediction stability, and DoF-based focusing accuracy.