Fuzzy C-Means Algorithm And Its Application In Image Segmentation

As a learning method of non-supervised, cluster analysis is an important technology in data mining. At the same time, it is an important analysis tool and method in data mining when we process the data. In the recent years, with the research in clustering technology, cluster analysis has become a research hotspot when we do the data analysis and information extraction in machine learning, data mining and also a lot of other areas.Clustering processes are always carried out in the condition with no pre-known knowledge, so the most research task is to solve that how to get the clustering result in this premises. The most research about clustering is focused on clustering algorithms, the main purpose is to produce practical algorithms with better performance. Up to now, many clustering algorithms have been presented, but these algorithms are only suited special problems and users. Furthermore, they are imperfect both theoretically and methodologically, even several faults. Optimizing deeply clustering algorithms will not only help to perfect its theory, but also its popularization and application. Fuzzy clustering analysis that introduces the theory of fuzzy sets, provides the capability that be used to deal with real data. And it has been widely used in many fields. In this thesis, we discussed typical fuzzy clustering algorithms. The advantages and disadvantages of these algorithms and the problems existing in these algorithms and the prospects of the fuzzy clustering algorithm are discussed.FCM clustering algorithm is one of the widely applied fuzzy algorithms at present. But FCM algorithm has some shortcomings. The FCM clustering algorithm is sensitive to the situation of initialization and easy to fall into the local minimum when iterating. In order to overcome shortcoming of FCM algorithm, a improved fuzzy c-means clustering algorithm is put forward. The basic idea of the algorithm is modified subjection value by adding weighted value and the optimal choice for parameter of clusters c based on cluster validity function. Then, the FCM based on SA-PSO is put forward. The algorithm can make use of the global optimization of PSO and the ability to jump out of local minimum of SA, so it can get a better clustering effect. To prove the availability of this improved FCM algorithm, we use the algorithm in image segmentation.Image segmentation is the process of detecting objects or interesting areas from input image,and it is an important step in object detection and recognition. Fuzzy clustering is an important branch of fuzzy set theory,and is widely applied in image segmentation. In this dissertation,the application of fuzzy clustering in image segmentation is studied. The main work of this dissertation is summarized as follows:(1) A fuzzy clustering algorithm is given which combines particle swarm optimization (PSO) and fuzzy c-means (FCM) clustering. This algorithm can reduce the influence of selection of the initial clustering centers value and the membership matrix elements on the algorithm convergence. The objection function for SA is set up according to FCM clustering,and the image segmentation algorithm based on SA and FCM clustering is implemented.(2)Another algorithm is also given in this thesis which combines simulated annealing(SA),particle swarm optimization(PSO) and fuzzy c-means(FCM) clustering,and apply it in image segmentation.Many researchers have done a lot of work and presented thousands of approaches on image segmentation. Unfortunately there is no universal method which could be used everywhere. There is even not an objective standard for evaluating segmenting algorithm. The clustering-based method is very important and wide-used in image segmentation. The most common method of clustering analysis for image segmentation is the FCM (Fuzzy C-means Clustering), which doesn't need setting any threshold, or get people involved. It is very significant for the automatization of image segmentation. Experiments of the proposed algorithm on synthetic test image and realistic image prove its validity and better performance against noise.