Research On Application Of Markov Random Field In Image Processing

Markov Random Field (MRF) theory had been widely used in the fields of computer vision and digital image processing. MRF provides a method which describes the correlated relation of images in two dimension system conveniently and directly. The proof of the equivalence of Gibbs distribution and MRF made rapid development of the application with MRF method. And, the conception of joint distribution in MRF gives the investigator a MRF model in Bayes system. The paper use the MRF theory in three representative application in digital image processing:Binary image restoration is widely used in many fields such as finger image preprocessing, text image restoration. The restoration algorithm look the original image as a MRF, and this is the prior knowledge of the estimation of Maximum a Posteriori (MAP).Which makes the difference is that how to calculate the MAP. Two traditional methods are used. One is a stochastic relaxation method, which is good at global optimality. The other is deterministic relaxation method, which is good at calculating speed. The paper provides a new changed simulated annealing method. The new method changes the J parameter dynamically during the iteration. The experience results proof that the new method can restore the dirty images, which is added with Gauss noise, with higher speed than SA algorithm. Image segmentation is a vital phase of image analysis. The method based on MRF takes the parameter of Gibbs as a representation of the texture character. So it is not affected by the noises. The paper provided a new method named pseudo-stochastic relaxation which is different from the SA, Gibbs sample and Iterated Conditional Mode (ICM).The first phase of the algorithm is similar to the stochastic techniques. The second phase is thus similar to the deterministic algorithms and it converges to a local minimum. Then the segmentation result is provided. The experiments have proved that the new algorithm be good at both global optimum and calculating speed. The single-frame object detection algorithm detects the different character between the object and the background and use the information in a single image to separate the object and the background. The energy function in Gibbs distribution can show the local texture character in the imaged. The new method finds where the object will be at first. Then we can separated the object from chosen area.