Abstract: Tobacco leaf disease and insect spot segmentation is an important guarantee to improve the quality of tobacco leaves. The traditional segmentation method falls short of the desired accuracy and efficiency. Aimed at this problem, a method of Otsu combining morphological and wavelet transform to divide tobacco leaf spot is presented. Firstly, the background area is processed by mathematical morphology in Otsu to get the image of tobacco leaf. Then the leaf image is decomposed by wavelet coefficients, after which the decomposed image is to be reconstructed with low-frequency so as to undo noise effects. Finally, the disease spots will emerge after the quadratic segmentation of the image in Otsu. Thanks to the morphological opening-and-closing operation that can extract the brightness details of the image, thus this method can effectively reduce the impacts from background onto the disease-spot area, which is to improve the accuracy and efficiency of segmentation. Meanwhile, segmentation precision can be further raised by multi-scale wavelet resolving to overcome the effects from redundancy and noise. The images of different sorts of tobacco leaf spots experimented, the results show that Otsu can effectively segment the tobacco disease spot, and operate on the lesions in other plants.

Key words: morphology, wavelet transform, Otsu method, tobacco disease spot, image segmentation