Medical Image Segmentation Based On Improved Contourlet Transform And Markov Field

BackgroundMedical image segmentation is of the great significance to image analysis and clinical diagnosis.In clinical application of medical image,medical image segmentation is the basis for 3D reconstruction,and medical image processing system can be clinically Practice and the basis of application.The rapid and accurate segmentation of medical image can not only meet the requirements of clinicians to complete the surgical plan,but also to ensure the accuracy of medical image imaging system in clinical application,and ultimately guide the follow-up work of clinicians.ObjectiveThe algorithm proposed in this paper is a new segmentation algorithm for medical image images combined with non-subsampled Contourlet transform and Markov field modeling.Firstly,the multidimensional direction coefficients of the image are obtained by using the Contourlet transform.Then the Contourlet transform direction is modeled by the Markov random field.Finally,the medical image is segmented by the maximum a posteriori probability criterion.The final results show that the algorithm can not only describe some details of medical image,but also make the segmentation result of medical image and the original image before segmentation maintain have a good consistency.MethodsThe paper has done some of the following work:Firstly,the frequency domain decomposition theory is studied,and the transform coefficients of the image in multi-direction are obtained by Contourlet transform of medical image.The non-subsampled Contourlet transform is used to deal with the image to be segmented,especially the image details.Sampling Contourlet transformations can better handle medical image images.Secondly,Markov modeling is carried out by Markov field,and the segmentation of medical image is realized by means of maximal a posteriori probability criterion,which can obtain more abundant image information and more sparse performance.The process is more concise.Thirdly,in order to evaluate the algorithm,the experimental scheme is designed in this paper.Through the comparative analysis of several groups of experiments,the five indexes given by the MICCAI conference group were used to evaluate the segmentation results of medical image.After a comprehensive analysis,the feasibility and practicability of the algorithm were explained.ConclusionThe innovation of this paper is to use the non-subsampled Contourlet transform method to remove the noise information before the segmentation of the medical image,and then combine the Markov field modeling,and achieve the image segmentation through the maximum a posteriori probability criterion,The detailed information of the image is described in detail,and the segmentation result is better than the original image,and it can be better applied in CT brain image and abdominal MR image data.