A Mothd For Segmentation Of Medical Images Based On The Finite Mixture Gauss Model

Medical image segmentation is an important application in the field of image segmentation. Thousands of methods have been put forward to medical image segmentation. Some use classical methods and others use new methods. This paper does the summary to the purpose and meanings that the medical image segmentation on the foundation of the existing research result; the classification and comparison to the existing medical image segmentation methods. This paper uses the brain MR image to experiment the object in numerous medical images, at the same time using the method on the rate statistics that is used in the segmentation of the brain MR image in iner common, namely a method according to the finite mixture Gauss model.This paper introduces the Markov random field and the Gibbs random fields first, although the Gibbs random field is not a kind of segmentation method, also does not the segmentation ability, it can be a check the model to imbed a certain and particular segmentation method, to attain the purpose of the influence segmentation result. Immediately after the paper researches the finite mixture model thoroughly cluding the deduce of the model and the application in the image segmentations .The result is that the finite mixture model while used for the image segmentation, considered the strength information of the pixel only, does not make use of any spatial information. Because the strength of the pixel distribute is independent, the classification of the images depends on the kind of histogram-based model completely, however, the image that has the same histogram-based can have the different structure completely. So that model has an intrinsic limitation in the image segmentation. This causes the FGM model to work only on well-defined images with low levels of noise and to affect the result of segmentation.In this paper, we present an improved segmentation mothd: A-prior spatial knowledge is incorporated into the FGM model through the Gibbs Random Field (GRF) theory. Forthmore the new mothd is also one with the model parameters estimated by EM mothd and inntialized by the tree-structure K-means mothd. Meanwhile, because of the feint in actule medical images, the actule mothd of Bias estimated is used in the new mothd.