Research On Spectral Clustering Algorithm For Image Segmentation

With the advent of the era of big data,the diversity and complexity of massive data has become the norm.Data mining technology extracts valid information from massive data and processes and analyzes it to achieve people's expected goals.Cluster analysis is one of the important research branches of data mining technology.As an effective method in cluster analysis,spectral clustering is based on the division of spectral graphs.The clustering problem is transformed into a graph division problem to achieve Clustering in a sample space of arbitrary shape.Because of its good clustering performance,spectral clustering is widely used in various fields of data mining.Among them,image segmentation has become one of the important application areas of spectral clustering algorithms.Image segmentation is a key step in the process of image data processing.The quality of image segmentation will directly affect the accuracy of image cognitive understanding.Therefore,the research around image segmentation has aroused widespread concern in academia and industry.The resolution of color images has gradually increased with the development of digital technology,which has led to an increase in image size.In order to solve the problems of spectral clustering algorithm in color image segmentation,such as large computational complexity,long processing time,and rough segmentation effect,this paper has conducted a series of researches on superpixel image preprocessing algorithms and spectral clustering image segmentation algorithms.The main contents are as follows:1.Aiming at the problem that the traditional spectral clustering algorithm uses pixels as a processing unit in image segmentation,the complexity of the operation is large.This paper introduces a superpixel algorithm to preprocess color images.The traditional superpixel SLIC algorithm requires artificially designed feature equalization parameters in the calculation process.In view of this deficiency,this paper introduces the idea of fuzzy mathematics to design a SLIC algorithm with adaptive equalization parameters,which can be based on the image during the calculation process.The specific conditions of the algorithm adaptively generate the feature equalization parameters required by the algorithm,thereby achieving a better segmentation effect in a low-cost operation time.The proposed algorithm is analyzed and verified through experiments.2.Aiming at the disadvantage that traditional Ncut algorithm needs to manually design parameters for clustering when solving the similarity matrix,this paper DGSV improves the similarity matrix in traditional Ncut algorithm,and improves the Ncut algorithm and adaptive equalization.The parameters of the SLIC algorithm are combined to generate a new superpixel-based spectral clustering color image segmentation algorithm.The algorithm in this paper introduces the superpixel algorithm to preprocess the image,which effectively reduces the calculation time of the subsequent spectral clustering algorithm in the subject segmentation process,and improves the processing effect of the traditional spectral clustering algorithm.In this paper,the proposed algorithm is experimentally verified and evaluated for two types of evaluation methods from different angles.Experiments show that the algorithm in this paper can fully consider the feature information and sample size of color image samples,improve the accuracy of image segmentation,shorten the processing time,and improve the operation efficiency of the algorithm.There are 25 figures,9 tables and 105 references in this paper.