Research On Image Segmentation Method Based On MAP-MRF

Digital image processing is a method of image processing that using a computer.Image is the foundation of human vision,and vision is one of the important measures about human perception in the world.Image segmentation is one of the basic problems in image processing.It separates an object into interest from an image by dividing it into several meaningful sub-regions or objects.Image segmentation is the pretreatment for image recognition.Correct segmentation will bring correct recognition results.Therefore,image segmentation is an important research technique in image processing.Researchers have proposed a variety of image segmentation algorithms,and statistical image segmentation methods have important significance for image processing.In this paper,we use the Maximum A Posteriori-Markov Random Field(MAP-MRF)framework and the Filter,Random Field,Maximum Entropy(FRAME)model to derive the Rudin-Osher-Fatemi(ROF)model and then introduce the process of improving the ROF model to obtain the Modified Mumford-Shah(MMS)model,we have reached an innovation using statistical methods to derive the MMS models.The MRF model is a statistical model that describes the relationship between image pixels.It can represent the pixel values with a certain probability distribution,then the pixel values can be represented by random variables.MAP is one of the most commonly optimization criteria in the MRF model,which can transform the maximization probability problem into a minimization function problem.To calculate our MMS model,we use the efficient algorithm of the fixed point iteration and the Split-Bregman algorithm.Finally we use the classical OTSU threshold image segmentation model and Canny edge detection algorithm to do numerical comparison experiments with the SB algorithm.The experimental results show that the SB algorithm has strong denoising capability,the detection edge is the clearest,and the segmentation result is the most accurate.