In order to accurately and quickly extract the focus position of pixels in the image sequence of the focus stacking 3D microscopic vision measurement, a focus measure algorithm based on the maximum gradient is proposed. Two band-pass mask operators and four high-pass mask operators are designed, and the maximum value of the intensity change of each pixel with respect to adjacent pixels is extracted from the band-pass signal and the high-pass signal to indicate the degree of focus, so as to improve the ability of focus measure function to extract gradient information. The adaptive gradient threshold segmentation algorithm improves the sensitivity of focus measure and reduces the influence of large fluctuations and sub-peaks of the focus measure curve. The experimental results show that, compared with the classical spatial focus measure algorithm, the focusing measure algorithm based on maximum gradient has significantly improved its robustness, sensitivity, unbiasedness and unimodality, and has good practicability, which can effectively meet the requirements of focusing measure of micro-nanometer scale 3D microscopic measurement of complex contours.