An Improved Method of Flotation Froth Image Segmentation Based on Watershed Transformation

Watershed transformation is an important method of image edge segmentation. The traditional watershed method is poor in dealing froth image mixed with different sizes of bubbles because single threshold value can’t eliminate noises in large bubbles on top and enhance the edges of small bubbles simultaneously. An improved method of froth image segmentation based on watershed transform was proposed in this paper. First, a high-low-hat transformation was done to froth image to increase bubble prominence. Then the markers of catchment basins were searched with different threshold values as to large, medium and small bubble objects. After that, the large and medium-sized catchment basins were morphologically marked and re-shaped. Finally, standard watershed transformation was executed on the target froth image. Experimental results show the proposed method is accurate in marking catchment basins with relatively good robust performance. In contrast to traditional methods, the proposed method is obviously more accurate in edge segmentation of froth image with an accuracy rate of over 80% and the accuracy rate is significantly higher in dealing with complex froth images.