Real-time auto-focus implementation

Summary Form Only Given. In the study of biological samples, it is often necessary to monitor and record cells in vitro via long term imaging. The images must maintain focus throughout the recording period. Many algorithms have previously been developed to quantify the sharpness of an image. We have developed an autofocus mechanism consisting of stepper motors and a camera under LabVIEW control. The actuators adjust the sample stage vertically in 0.16mm increments from the objective lens. To test the system a sequence of epithelial cell images was captured using a camera (Point Grey 14SSM-C) with a 60x (NA1.0) water-dipping objective lens. The images were then analysed off-line with a variety of different algorithms (MATLAB, Focus Measure, Said Pertuz, Apr 2010) to determine the most robust and accurate focusing method. The grey-level local variance (GLLV), Vollat's correlation-based (VOLA) and the thresholded gradient (GRAT) methods were the most accurate at obtaining the target focus position. They all possess a single maxima at the sharpest image. All algorithms had comparable processing speed (10 images per second) however many of the tested algorithms failed to focus on the target image. The best algorithm (GLLV algorithm, Pech-Pacheco JL et al. 2000) was then implemented in a real-time LabVIEW auto-focusing system. When the sharpness falls below a threshold the system repositions the objective to restore the focus. In summary, a highly controllable, rapid and accurate auto-focusing system has been developed that will be useful in long-term imaging of cell cultures.