Research On Image Segmentation Algorithm Based On Active Contour

Image segmentation is an indispensable part of image processing,target detection,pattern recognition,computer vision and machine vision.It plays an important role in medical image analysis and remote sensing image extraction.Since the 1990 s,the image segmentation method based on active contour is finding wider and wider application in the field of image segmentation,in view of the accuracy of image segmentation method in obtaining the boundary of the object,can achieve sub-pixel level,and the resulting contour is smooth and closed,therefore,the next image analysis and target detection provides a good foundation.This paper aims to explore the image segmentation model based on active contour,which is beneficial to the segmentation of images with uneven distribution of gray scale,but also has some limitations and problems.For example,the model is often sensitive to the setting of initial contour and the segmentation speed is slow.Based on the above considerations,this paper proposes two innovative points based on the original model:1.Improve the active contour model based on local fitting.This method is mainly when the curve evolution,the local area in some function value in the opposite direction,then the opposite function fitting function values of exchange,so that in the process of the evolution of the curve,the outline will outline than internal fitting values outside of the big or small fitting values,then the whole curve evolution will be along the boundary within or outside the boundary of the target,and don't not make the final evolution curve in the interior of the target,then it can be solved when the energy minimization into local optimal solution of the problem.This improved method not only retains the advantages of the traditional method,but also improves the robustness of the initial contour.2.Before the curve evolution is driven by the activity model based on local fitting,the average gray value of the local image is calculated in advance,that is,two functions for local fitting are defined in advance.Compared with the traditional function for local fitting,the difference is that the pre-fitting function proposed in this paper has nothing to do with the level set function and does not need to iterate every time when the function is updated.Therefore,compared with some traditional active contour models based on local area fitting,this improved model has the advantages of less computation and faster segmentation.In addition,because the initial horizontal set function of this model is set to a constant,it has good robustness to the initial contour.From the experimental results,it can be seen that the two improvement points proposed in this paper can achieve better segmentation results for images with uneven distribution of gray scale,and can also achieve good segmentation results for images with weak edges or noise.Compared with the traditional local fitting model,the first of the improved model is better to solve the traditional active contour model of the selection of initial contour is relatively sensitive problem,while the second improved model in view of the traditional active contour model segmentation slower shortcomings,sped up the segmentation speed,improve the segmentation efficiency,and enhance the robustness of the initial contour.