The Image Segmentation Method And Research Based On Markov Random Field

With the development of science and technology,Image technology have got hugeattention and considerable development.Among them, image segmentation is one of the keytechnologies of digital image processing,and one of the hotspot fields of researchingimages.Image segmentation is a critical step between the image processing and the imageanalysis.The result of image segmentation has a direct impact on the progress of the follow-up work.,Ideal segmentation results are based on the priori understanding knowledge.This isthe difficulty of image segmentation.This paper is used to solve this problem based on thesingle-scale Markov random field image segmentation.On the basis of previous work,it has adeep discussion and research on how to construct Markov random field model and how tochoose a fast segmentation method.Firstly, on the part of image preprocessing,it used to denoise the image with the medianfilter to make the image more clear and then K-Means algorithm is used for initialsegmentation.Secondly,the single-scale Markov random field model theory is discussed,itmainly researches Bayes estimate, expectation-maximization algorithm of parameterestimation in the Characteristic field and Maximum A Posteriori Estimation Theory in theMarkov random field.In the construction of Markov random field model,label model fieldand characteristic field are discussed,After comparing the advantages and disadvantages ofthem,Multi-level logistic model is used in the label model field and finite Gauss mixturedmodel is used in the characteristic field.According to the Markov-Gibbs equivalence, itconvers the image labeling problem to the energy problem.Finaly,In the segmentationalgorithm,iterative condition model is improved in this paper.The stable point is labeled byan array and each iteration are just to solve the unstable points.The experimental results showthat the algorithm reduces the amount of computation and improves the computationalefficiency without influencing the segmentation effect.This article implements the relevant algorithm in the Microsoft Visual Studio2008programming environment.The test results indicates that Markov random field imagesegmentation has a good result after the initial segmentation of K-Meansalgorithm.Otherwise,Using the improved ICM algorithm can improve the time efficiency.