A new auto-focusing function for microscope measurement is presented after the research of imaging and focusing principles in order to,in automatic microscope measurement techniques,not only enhance the accuracy of the focus value but also suppress noise. At first,the measuring edges are detected by the improved SUSAN algorithm and lens is moved. When the total of edge points detected is least,then the best-focused lens position is found. We tests in the same condition for 30 times,and experimental results indicate that standard deviation of focusing position in coarse focusing and in fine focusing is 0.000 and 0.0011 respectively,focusing in Gauss noise case is 0.0126,and focusing in salt-and-pepper noise case is 0.0033. BW is accurate in focusing degree,unimodal,sensitive to focusing position and insensitive to noise compared with traditional focusing functions. BW is one of the superexcellent focusing evaluation functions with simple principles,which is a computationally efficient alternative for microstructure measurement.