Strong Parameter Adaptive Spectral Clustering Image Segmentation Based On SLIC Algorithm

With the development of information technology,the era of big data has come.In the field of data mining,clustering is a crucial method of data analysis.Cluster analysis is also a branch of multivariate statistical analysis and unsupervised machine learning,which is widely used in various fields.On some more complex or challenging clustering problems,spectral clustering is an outstanding tool,the image segmentations are impled by using graph theory,the method is to divide the set of points in the data set by solving the eigenvector corresponding to the eigenvalue of the similarity matrix in the data set,the data clustering problem is transformed into an image segmentation problem,and the purpose of clustering is realized by the optimal segmentation of images.Image segmentation is an significant step in processing image data,so it is vital to improve the performance of image segmentation.Nowadays,it is greatly concerned about the research on image segmentation in academic community,in the process of image segmentation,the problems of large amount of computation and complex degree of computation are still exist,When the traditional algorithm of spectral clustering is used for image segmentation,the performance of image segmentation will be seriously affected by the limitation of the algorithm,however,this problem can be relieved by introducing the concept of superpixel and its related algorithms.In this thesis,a Simple Linear Iterative spectral Clustering algorithm based on superpixel is introduced,which is called SLIC a Igorithm,in this theoretical context,a powered parameter adaptive spectral clustering algorithm based on SLIC algorithm is proposed,the algorithm is realized by the image segmentation examples and the related research is carried out.This thesis firstly introduces the research status and fundamental principle of spectral clustering algorithm,based on the background of spectral clustering algorithm,Ncut algorithm,a classical traditional algorithm in spectral clustering,is introduced,furthermore,the basic principle,division rules and algorithm steps of Ncut algorithm are introduced,which is used to achieve image segmentation;then,in order to solve the problem of high computational complexity of classical spectral clustering algorithm in image segmentation,SLIC algorithm is introduced and used to preprocess the image pixels;finally,an improved measure is proposed which is directed at the similarity function of traditional Ncut algorithm in spectral clustering,by introducing a powered parameters can improve the clustering effect,at the same time,the adaptive parameter is defined,which make the parameter of the function is setted in the algorithm instead of setting manually,the performance of spectral clustering algorithm is greatly improved,combining the SLIC algorithm with the improved Ncut algorithm,a powered parameter adaptive spectral clustering algorithm based on the SLIC algorithm is proposed,the effectiveness and feasibility of the algorithm are verified by examples of image segmentation and the corresponding experimental evaluation criteria.The innovation of this thesis is that an improved powered parameter adaptive spectral clustering algorithm is proposed,which is greatly improved the performance of the algorithm,to give sufficient consideration to the image color,spatial distance and other characteristic information,the powered parameter adaptive spectral clustering algorithm based on SLIC algorithm is used to achieve image segmentation,and the image segmentation effect is remarkable.