Abstract: In view of the complexity of plant leaves images, combining with the characteristics of plant color leaves images,a color image segmentation method based on color differences was presented. Firstly,the images of the diseased plant leaves were collected,use GrabCut algorithm to segment the complex field environment background. Then,use median filtering and image sharpening method to preserve the image of disease area and edge detail. Next,the processed images are converted to Lab color space and YUV color space which based on physiological characteristics. Combined with Otsu method,the images of Lab grayscale and single channel grayscale of YUV color space are subdivided respectively. Finally,the two subdivision of the two figures and operations,its conversion to RGB color space,you can get the final segmentation results. The method was used to segment the common color images of leaves diseases plants such as soybean,corn,rape and cucumber. The experimental results showed that this method can obtain more accurate segmentation,Edge detail protection and segmentation efficiency also have good results.

Key words: plant, leaves, color image segmentation, color difference, Otsu