Research On Cervical Cell Image Segmentation Based On GVF Snake Model

Image segmentation technology has always taken an important role in the field of image processing and computer vision,and the algorithm of image segmentation has also been a hot research area for scholars both domestic and overseas.Among many mature image segmentation algorithms,object contour extraction based on Active Contour Model has gotten extensive attention,which combines the characteristics of the image itself with prior knowledge.Gradient Vector Flow Model is one of the Parametric Active Contour Models,which can provide a good solution for most image segmentation problems.Firstly,this paper introduces the background and significance of image segmentation technology,and summarizes the research status of cell image segmentation methods.Secondly,we elaborate on the fundamental principle and algorithm of traditional Snake model.Based on that,this paper introduces the Balloon Snake model and GVF Snake model,and compares the segmentation effect of these two improved Snake models through simulation experiments.Moreover,the method of setting the initial contour of GVF Snake model is discussed in detail.The Otsu method and the edge improved Otsu method are proposed to improve the accuracy of setting the initial contour.The aim is to obtain better segmentation results and faster convergence speed.Finally,cervical cell image has the characteristics of regular shape,centralized nucleus,small size and opacity.GVF Snake model is applied to the segmentation of cervical cell image.The main works of this paper are listed as follows:1.Based on the traditional Snake model,its basic principle and algorithm are discussed and simulated on the MATLAB platform.Then we describe the advantages and disadvantages of the traditional Snake model.Aiming at various improved Parametric Active Contour Models,the principle and algorithm of Balloon Snake model and GVF Snake model are emphatically analyzed,and we compare their merits and demerits through simulation experiments.The simulation experiments verify the advantages of GVF Snake model.It is also proved that GVF Snake model can greatly improve the sensitivity of the initial contour position.GVF Snake model can significantly enhance the accuracy of segmentation and the convergence results of the algorithm.2.When we research the initial contour setting of GVF Snake model,three methods of initial contour setting are described and compared particularly: Wavelet Transform,Watershed Algorithm and edge detection operator.Among them,the Canny operator and LoG operator are compared and analyzed emphatically.LoG operator is selected as the contour extraction method of GVF Snake model in this paper,and the accurate and closed initial contour curve of the target object is obtained by MATLAB experiment.3.The GVF Snake model is used to segment the cytoplasm and nucleus of cervical cell image.We minutely introduce the steps of image preprocessing,then the gray scale and denoising methods are compared by experiment.In this way,we choose the most suitable algorithm for cell image,which is very helpful for the subsequent segmentation using GVF Snake model.Before extracting the initial contour needed by the model,we propose a method of rough nucleus segmentation which can use gradient edge information to improve global threshold.The interested target region can roughly segmented before extracting the initial contour.The results of rough segmentation are also quantitatively analyzed.We implement the cytoplasm and nucleus of cervical cell image segmentation by GVF Snake model on MATLAB platform.The experimental results show that GVF Snake model can keep the contour of the target object precisely,and obtain the contour curve of the cytoplasm and nucleus of cervical cell accurately.The segmentation results accord with the expected results.