Research On Active Contour Segmentation Method And Its Application On Medical Image

Medical image segmentation is the basis of high-level clinical diagnosis and pathological research.Due to the uneven gray scale,the low contrast between the target and the background,and the noise in the medical image,it is easy to cause the target contour to be blurred,which adds difficulty to the segmentation of target of the medical image.Active Contour Model(ACM)is a segmentation method based on curve deformation.The evolution of closed curve is controlled by the speed function derived from minimizing energy functional.Active Contour Model is widely used in medical image segmentation field because of its good performance.We focus on the improvement of the external force field of the active contour model.To address the limitations of the Generalized Gradient Vector Flow(GGVF)model and the Virtual Electric Field(VEF)model,we propose a corresponding solution and the main work includes:We propose Robust Generalized Gradient Vector Flow(RGGVF)model.The model has three improvements on the basis of the original GGVF model:(1)A bilateral filter is incorporated to build a novel edge map.It can not only suppress noise but also retain some important edge information,and possesses a better effect on reducing noise interference.(2)The Laplacian operator is replaced with the divergence operator.This changes the diffusion process of gradient vector from edge areas to homogeneous areas and the performance of convergence to concavities can be improved.(3)An orientation constraint function is employed.This makes the diffusion at the edges carried out in a direction perpendicular to the gradient without blurring the boundary.Accordingly,the weak edge can be protected effectively.In this thesis,experimental results in qualitative and quantitative metrics are used to comprehensively evaluate the algorithm.The experimental results in qualitative demonstrate that the algorithm outperforms the competing ones regarding noise robustness,weak edge preserving and deep concavity convergence,and has a good segmentation effect on general medical images.Experiments results in quantitative show that the RGGVF model increases about 0.3%,15.6%,and 8.7% in precision,recall,and F measure compared to the traditional GGVF model.We propose Edge Preserving Virtual Electric Field(EPVEF)model.A vector value function that can change the vector size is employed to reducing the influence of pixels in homogeneous regions on edge pixels and prevent edges from being over smoothed.As a result,the edge protection ability of the model can be improved.The experimental results of segmentation experiments on synthetic images and medical images show that the RGGVF model not only has the advantages of large capture range and fast calculation efficiency,but also has better edge protection ability,and is more suitable for segmentation of target in neighboring organizations of medical images.