Autofocusing algorithm comparison in bright field microscopy for automatic vision aided cell micromanipulation

Abstract

Autofocusing is an essential technique in many machine vision aided microscopy application. This paper presents a comparison study of 6 autofocusing algorithms under bright field illumination: a) Normalized Variance (VAR), b) Tenengrad Gradient (TEN), c) DB06 wavelet filter (DB06), d) Fast Fourier Transform (FFT), e) Standard Deviation (STD) and f) Sum Modulus Difference (SMD). In the study, all the 6 algorithms are integrated with the exhaustive search technique and implemented using LabVIEW on a Pentium 4 desktop computer. A total of 2,204 microscope images of a micropipette tip are acquired at different microscope objective positions controlled by a high precision stepper motor under 2.8X magnification, are used to evaluate the performance of the algorithms in terms of processing speed, accuracy, consistency, sensitivity to image size and sensitivity to movement step resolution. It can be concluded that VAR and STD perform well in all performance measures.