Abstract: In order to address the segmentation errors of the traditional watershed algorithm in flotation foam image segmentation,this study integrates the SoftEdge model with an improved watershed algorithm. First,Gaussian low-pass filtering is applied to denoise the foam images. Then,the SoftEdge model is used to extract soft edges,thereby reducing the interference of light noise in edge detection. Subsequently,a watershed algorithm optimized with foreground-background markers is adopted. By accurately extracting these markers,the algorithm performs segmentation only within predefined regions,which significantly reduces segmentation errors. The results demonstrate that this method avoids reliance on prior knowledge and complex parameter settings,while substantially improving segmentation accuracy. 

Key words: image segmentation of flotation froth,SoftEdge model,improved watershed algorithm,foreground and background marking techniques