A novel focus measure algorithm for three-dimensional microscopic vision measurement based on focus stacking

Abstract

The three-dimensional (3D) microscopic vision measurement methods based on focus stacking are widely used because of their outstanding advantages such as noncontact, high accuracy, high flexibility, and low cost. On-site focus stacking 3D microscopic measurement is susceptible to environmental disturbances, surface dirt, illumination occlusion, and the shape of the object itself. This leads to the difficulty of obtaining the focus position accurately, stably, and quickly with the existing focus measure function. To improve the ability to extract the focus signals from the image, high-pass mask operators and band-pass mask operators are designed. A new high-pass and band-pass energy weighting focus measure function with rotation invariance is proposed. A focus stacking microscopic measurement system is constructed to validate the proposed focus measure function. Six classical and latest focus measure functions are selected for the quantitative performance comparison experiments. The experimental results show that the proposed algorithm has the best performance in accuracy, sensitivity, unimodality, and monotonic interval range. This measure function has a high operational efficiency, nearly three times the computational efficiency of the frequency domain measure function. The measure function also drastically reduces the evaluation curve fluctuation, peak collapse, and other anomalies, and shows good robustness.