Method For Medical Image Segmentation Based On Active Contour Model

The purpose of medical image segmentation is dividing medical image into several disjoint "connected" area, and make segmentation results as much as possible close to the anatomical structure. This can provide a reliable basis for clinical diagnosis and pathological study. Due to complexity of the human anatomy, tissues and organs shape irregularity and individual differences and other factors, accurate medical image segmentation become one of the most challenging problems in medical image analysis.Recently, image segmentation algorithm based on level set active contour models with its variety of forms, flexible structure and superior performance, received extensive attention from scholars. In this paper, we do more in-depth study for this type of image segmentation algorithm.Firstly, we have introduced active contour models based on the traditional level set image segmentation method, such as geometric active contour model, without re-initialize model and Chan-Vese model.Secondly, we have in-depth studied image segmentation method based on the binary level set active contour models. The binary level set active contour model uses the binary level set function instead of the traditional level set function, thus avoiding the level set function is reinitialized to the signed distance function, greatly improve the computational efficiency. At the same time, the binary level set active contour models also maintain the ability of the automatic processing of contour topology changes. In this paper, we has introduced LBF (local binary fitting) model based on the binary level set function proposed and the LFI (local fitting image) model based on the binary level set function proposed. Against losing of the progressive evolution curve of the LFI model, this paper presents an improved model. In this model, adsorption factor is leaded into LFI model as a new force of the curve, in this force, evolution curve outskirts the target boundary steeply and orderly to reduce the segmentation affect of the image background to the target, makes the model segmented Against the Target boundary, therefore achieved the desired segmentation results.