Research On Interactive Color Image Segmentation Algorithm Based On Superpixel And Spectral Clustering

With the development of science and technology,how to deal with the received color images for further analysis and research has become an urgent problem in image engineering.Image segmentation is the basis of image processing,directly affect the results of image analysis.The image segmentation based on clustering has always been a hotspot,especially the spectral clustering algorithm,because of it has no restriction on the distribution of sample space and can converge to the global optimal solution,to obtain good results in image segmentation.However,the traditional spectral clustering algorithm is based on the theory of graphs,still has many unresolved problems in massive data operation and similarity structure.An interactive color image segmentation algorithm based on super pixel and spectral clustering is researched in this paper.This method firstly introduces the super pixels to reflect the integrity of the subsequent image segmentation results,then some improvements were made by super pixels image,including the sampling strategy,similarity measure construction and the tag information using in spectral clustering algorithm.This paper mainly aims at the following points:(1)The spectral clustering algorithm is based on the spectrum theory,it has not yet put forward a good solution in how to deal with large-scale data and how to select the effective features of the image.Aiming at this problem,a Nystrom spectral clustering algorithm based on super pixel and region feature is proposed.In this algorithm,the spectral clustering algorithm is combined with the super pixel,and the regional feature selection is introduced.The strategy of sampling sampling is designed in the Nystrom algorithm,which effectively improves the efficiency of the operation and solves the problem that the similarity matrix memory is easy to overflow,To improve the stability of the algorithm.The experimental results show that this algorithm is superior to the traditional FCM algorithm and Nystrom algorithm in segmentation result and segmentation efficiency.(2)The selection of the sample points directly affects the stability and validity of the Nystrom algorithm.In this paper,a Nystroom spectral clustering algorithm based on(3)artificial sampling is proposed by using the user-given sampling mark.In the algorithm,the combination of sampling and manual sampling method,through the artificial guidance of sampling selection,not only makes the sampling points more dispersed,and more conducive to select the information in some small targets,get accurate segmentation effect.(4)The construction of similarity matrix is a key step in spectral clustering algorithm.In this paper,based on the robustness of image area integrity and similarity measure,the super pixel and fuzzy theory are introduced,and a semi-supervised color image segmentation with superpixels and spectral clustering is presented.Firstly,initial partitioning is performed by the superpixels method.Then the semi-supervised fuzzy similarity measure among the initial partition regions is constructed by utilizing the few label information.Finally,the similarity matrix of the initial partition regions is produced by this similarity measure,and these regions are grouped by normalized cut criterion.Because of the introduction of the label information and fuzzy theory,the experimental results show that the segmentation accuracy and computational complexity of the proposed algorithm paper are improved substantially.