A Segmentation Method Based on Gray-Scale Morphological Filter and Watershed Algorithm for Touching Objects Image

Abstract

In the process of intellectualized visional testing products, valid image information is often collected to analysis for objective parameters such as quantity, size, and shape and so on. Because there is phenomenon of touch, the objects needs to be segmented accurately. A watershed segmentation algorithm based on distance image using grayscale morphological pretreating is presented in this essay. First, distance transform is applied to the binary image. And then opening operations of gray-scale morphology are applied to remove small light details in distance image before applying watershed algorithm. Thus oversegmentation was controlled and touching objects were segmented precisely. The result of the experiments shows that touching objects can be segmented with the accuracy near 100% with this method, that the shape of the object is not to be changed, which can meet the demand of analysis of object characteristics in intellectualized visional testing process, and that it has a practical value.