Research On Image Segmentation Algorithms For Robust Asymmetric Finite Gaussian Mixture Models

In the fields of machine learning and data mining,model-based clustering analysis is more important,and the commonly used model for clustering analysis is a finite hybrid model.This model has a great influence on the development of clustering analysis.The quality of modeling also directly influences the effectiveness of clustering analysis.In this paper,based on the theory of asymmetric distribution and finite mixed model,the image segmentation algorithm based on asymmetric finite mixed model is studied,and the advantages and disadvantages of different segmentation strategies are discussed.The mechanism of priori representation of cerebral magnetic resonance(MRI)images is studied,and the theory and application framework of asymmetric finite mixing model are constructed.The main work and research results of this paper are as follows:(1)From the aspects of distribution selection,the introduction of Markov's airport,and the establishment of asymmetric distribution,the image segmentation method based on Gaussian hybrid model is deeply studied and analyzed.(2)From the two aspects of de-noising and segmentation strategy and de-deflecting field and segmentation strategy,the paper discusses the advantages and disadvantages of the segmentation strategy of cerebral magnetic resonance image.The mechanism of prior representation of MRI images is fully explored from spatial constraints,prior constraints,boundary constraints,offset field models.(3)A priori probability distribution based on data boundary constraints is established,and a segmentation framework with spatial and boundary constraints is constructed;(4)Based on the segmentation framework of asymmetric finite mixing model,a partial displacement field estimation model is introduced,and an image segmentation method based on asymmetric finite mixing model is implemented.