Abstract: In order to study the effects of surface microstructure indexes on tobacco leaf segmentation, the B2F and C3F flue-cured tobacco leaves of cv. Yunyan 87 were equally divided into 10 sections longitudinally after removing the petioles. The cell areas, cell perimeters, cell morphological factors, cell densities, stomatal densities and stomatal indexes of the different sections were quantitatively analyzed with scanning electron microscope and PHOTOSHOP software. Regression analysis was conducted to study the changes of the surface microstructure indexes for the different sections of tobacco leaves from two stalk positions, and the optimal segmentation ratio and sensory quality differences of tobacco leaves from two stalk positions were analyzed by Fisher best division and sensory evaluation. The results showed that: 1) The surface microstructure of the different segments of tobacco leaves were different. From the base to tip of the leaf, cell area, perimeter and morphological factors presented a parabolic trend that increased first and then decreased, while the cell density and stomatal density showed another parabolic trend that decreased first and then increased. The stomatal index increased linearly in general. When the same segment of B2F and C3F leaves were compared, cell area, cell circumference, stomatal density and stomatal index of B2F were higher than those of C3F, and cell morphological factor and cell density were reverse. 2) Based on the data of standardized microstructure indexes, Fisher optimal segmentation method was used to determine the optimal segmentation ratios (base: middle: tip) for B2F and C3F tobacco leaves of Yunyan 87, and the ratio for B2F was 20%: 50%: 30% and that for C3F was 30%: 50%: 20%. 3) The sensory results indicated that the overall sensory quality of the different segments of tobacco leaves differed significantly, which suggested that the surface microstructure of tobacco leaves combined with Fisher optimal segmentation method could effectively divide tobacco leaves into segments with great differences.