Neutrosophic Fuzzy Clustering Image Segmentation Algorithm Based On Wavelet And Spatial Information

Image processing is an important means of human visual continuity.In the research and application of image,people may be interested in some parts of the image,which correspond to the specific or special features of the image area,that is,the target and background,that is,the segmentation process.Although there are numerous research results on image segmentation,there is still no universal segmentation method.Therefore,the significance of image segmentation is extraordinary,which needs in-depth study and extensive expansion,and more researchers need to concentrate on analysis and research.Fuzzy clustering is a very useful technique in image processing.Each sample in fuzzy clustering no longer belongs to a certain class,but belongs to each class with a certain degree of membership.That is,through the fuzzy clustering analysis,we can get the uncertainty degree of the samples,and establish the uncertainty description of the samples for the categories,which can more accurately reflect the display world.Fuzzy clustering can not only extract features directly from the original data,but also optimize and reduce the dimension of the features that have been obtained,so as to avoid "dimension disaster".The application of fuzzy clustering algorithm can improve the efficiency of image segmentation.The main contents of this paper include the following three aspects:(1)Due to the poor effect of FCM in the presence of noise,outliers or other image artifacts,the idea of neutrosophic set is introduced,and the improvement measures of neutral fuzzy clustering algorithm are proposed.The clustering center of fuzzy items is defined according to the distribution of membership data of deterministic subset,and the weightcoefficients of deterministic class,fuzzy class and noise class are adapted,which effectively reduces the convergence time and improves the clustering efficiency.(2)Aiming at the problem that the segmentation accuracy of neutrosophic mean fuzzy clustering algorithm is not high and vulnerable to external objective factors,we introduce the idea of bias field estimation to correct the classification of sample points,which can effectively improve the accuracy of segmentation.(3)In order to balance noise removal and preserve image details,wavelet threshold noise removal is introduced for preprocessing,which can retain more details.At the same time,we also use the non-local spatial information processing algorithm,which has obvious effect on anti-noise.