The invention discloses a global optimization image segmentation method based on seed point automatic selection. The method differs from the traditional method in the following aspects. The method doesn't depend on the selection of the starting point of the seed and comprises the following steps: first, directly according to the target pixel threshold characteristics, performing global preliminary screening to all pixels of an image; then according to the row-and-column positions of all the pixels after the preliminary screening, searching and generating a target pixel position set; then according to the horizontal block direction threshold and the vertical direction block threshold, performing automatic division to the multiple target areas which can automatically and continuously generate multiple closed target areas at the same time, and makes up for the defects of a traditional image area growing method relying on the seed point selection; and finally, according to the column threshold of the valid target areas, determining the validity of the target areas so as to determine the final target segmentation areas.