Research On Image Processing Algorithms Based On The MRF Model And The Improved DCGAN Model

Image processing plays an important role in image engineering.This dissertation focuses on image segmentation and image fusion based on the Bayesian theory.Image segmentation is a key technique in image processing.In this paper,a Monte Carlo segmentation algorithm based on the MRF(Markov Random Field)image model is proposed.In order to avoid the over-dependence of the algorithm on the initial value and overcome the shortcomings of the existing iterative algorithm in local optimal solution,the model parameters are initialized randomly.Firstly,the MRF model can make full use of the neighborhood relationship of pixel space to obtain the image data field information.Secondly,according to the Bayesian theory,the prior knowledge of the image is transformed into the prior distribution model.Finally,Monte Carlo segmentation algorithm is used to iterate until the maximum posterior probability is reached,and then the distribution of image labels is obtained,that is,the process of image segmentation is completed.Due to the limited depth of field of optical lens imaging system,multi-focus image fusion is also a major research topic of image processing,which is widely used in military,remote sensing and medicine.Based on the deep learning theory,a multi-focus image fusion algorithm is proposed,which combines deep convolutional network and adversation generation network.Compared with traditional image fusion based on the multi-scale transformation,it has a better fusion effect.Experiments show that the monte carlo algorithm can overcome the disadvantage of traditional iterative algorithm falling into local optimum,and segment the image more perfectly and meticulously,effectively realize the accuracy of segmentation,and improve the iterative speed of image segmentation.The improved DCGAN(Deep Convolution Generative Adversarial Network)algorithm can overcome the shortcomings of the traditional algorithm,and has a better fusion effect.