Abstract: In order to solve the over-segmentation problem in traditional watershed segmentation algorithm in slime flotation froth images,this paper proposes an improved watershed segmentation algorithm method which is based on adaptive maxima-marker extraction.Firstly,it carries on Gaussian filtering to flotation froth image,then,the markers are adaptively extracted from the bubble image by using the maximum expansion technology based on the morphology,after that the markers are used to modify the gradient image.Finally,the revised gradient image is processed with watershed segmentation algorithm.The experimental results show that improved algorithm solved some problems,for example,we must grasp a priori knowledge before extracting the markers,the segmentation process is complicated and so on,and the improved algorithm can select parameter in a more reasonable way,and get a more accurate segmentation result.