A New FCM-Based Algorithm of Hydrophobic Image Segmentation

In order to effectively distinguish the objects from the background, three transition region feature models were taken into account in this paper, including gradient, local complex and local fuzzy variance. Firstly, three transition region feature models were used to distinguish the transition region and the smooth region. Secondly, experimental results signified that every feature models had effects on extracting the objects from the background and each model had some disadvantages on extracting droplets (or watermarks) from hydrophobic images. Finally, these three models were combined together to form the feature domain for fuzzy C-means clustering (FCM), and then the FCM was applied in segmentation of hydrophobic images. Experimental results illustrated that the proposed algorithm has a great application in segmentation of hydrophobic images, for its efficiency in extracting shape information of droplets (or watermarks).