Segmentation of Somatic Cells Based on Cloud Model

Mastitis is the major cause of loss in dairy farming. Somatic cells are one of most important standards to detect this infection. This work applies to a blended segmentation method for counting them in bovine milk images. Firstly, it uses cloud model to threshold cell images; secondly, watershed is used to reconstruct distance image in order to obtain the initial segmentation result; finally, according to the real sense of region similarity of human vision, an region comprehensive similarity criterion is definedconsidering gray distance, variance and edge information, which is used tomerge the pre-segmentation regions to a final segmentation result. The experimental results have been obtained using a large set of images from different sources. A performance comparison between the manual counting and the proposed method has indicated that the later one is a promising solution to automate systems for detection of bovine mastitis via optical microscopy.