Research On Robust Fuzzy Clustering Segmentation Algorithm With High Performance

Image segmentation is one of the basic problems in image processing and computer vision.It is the key to realize image processing,image analysis and image understanding.The quality of image segmentation has an important influence on image understanding,and the segmented region can be used as the target object for subsequent feature extraction.At present,the research of image segmentation involves computer science,pattern recognition,machine learning,psychology,artificial intelligence,communication transmission and so on,its new results promote the development of the related areas greatly.There are various kinds of image segmentation method and the fuzzy C means clustering algorithm(Fuzzy C-means clustering algorithm,FCM)is the most widely used algorithm.Traditional fuzzy C-means clustering algorithm based on fuzzy sets and traditional Euclidean distance objective function,and the algorithm does not consider the spatial information of pixels,vulnerable to the impact of outliers,deficiencies in the quality of classification,and the algorithm for more sensitive to noise in the image,poor clustering effect.At present,many scholars have put forward improved algorithm to improve the robustness of the algorithm,the algorithm has better segmentation effect,such as the kernel fuzzy C-means clustering algorithm(Kernel fuzzy C-means clustering algorithm,KFCM),to map the input data space to a high dimensional feature space for clustering by using this method,a better clustering effect,at the same time the convergence speed is fast.Fuzzy C-means clustering algorithm based on spatial information(Spatial fuzzy C-means clustering algorithm,FCM S)and nuclear spatial information fuzzy C-means clustering algorithm(Kernel spatial fuzzy C-means clustering algorithm,KFCM_S),introduce the image information to the neighborhood clustering objective function,each iteration calculation of membership degree and cluster center integration sample neighborhood information,eliminate noise.In order to improve the robustness of noise interference image,this paper propose a kernel space hidden Markov random field FCM algorithm,kernel space adaptive non-local mean robust segmentation algorithm and FCM algorithm based on feature selection.The main work of this paper is as follows:1.Introduce fuzzy set theory,FCM algorithm,KFCM algorithm,analysis of FCM clustering algorithm for image segmentation and clustering errors occur,the error caused by the division of the image segmentation is not clear enough.Combined with hidden markov random field model(Hidden Markov random field model,HMRF),and the algorithm is generalized to kernel space,proposed kernel space hidden Markov FCM algorithm to describe the spatial information among pixels with probability,and the membership function for hidden markov optimization by introducing a prior probability function,so as to obtain the optimal solution approximation the membership degree and the maximum to obtain segmentation markers,fully considering the randomness of the image,the pixel clustering is more accurate and better robustness.2.The non-local neighborhood information embedding FCM segmentation algorithm,improve the anti noise performance of the algorithm,however,due to the non-local mean filter parameters are fixed,the segmentation algorithm for different intensity of image noise is lack of universal.In order to improve the performance of the algorithm,this paper proposes an adaptive non-local mean fuzzy C clustering algorithm based on adaptive filtering algorithm.Compared with FCM,KFCM,FCM_S,KFCM S evaluation index PSNR value,the algorithm in this paper is improved by at least 1-2db.In this paper,the result of image clustering segmentation is more stable,the consistency of the segmentation region is better,and the robustness is stronger,meet the needs of noise image segmentation.3.According to the conventional fuzzy C means clustering algorithm is an unsupervised method,no pre labeled categories for the training set,can not accurately determine what data is useful,what data is useless,so can not get accurate classification.Combining the concept of feature saliency and the method of tag selection,a FCM algorithm based on feature selection is proposed.The feature selection method of image data can effectively improve the classification performance of the algorithm,so as to improve the anti noise performance of the algorithm.