Inhomogeneity Image Segmentation Research Based On Active Contour Model

Image segmentation is the process of dividing the image into different regions which is similar or consistent,according to different features.It is a very important pretreament in the field of pattern recognition and computer vision.It plays a key role in the image feature extraction and research.With the development of new theories and methods and the continuous improvement of the original models,many segmentation algorithms have been developed in recent years.However,there is no a general method to segment all images because of the wide variety of images and the complexity of application.The active contour models(ACMs)based on the partial differential equation is a kind of emerging image processing model,which is developed by relying on curve evolution theory.Compared with the traditional image segmentation algorithms,it has the advantages of stronger adaptability and higher segmentation precision.This paper focuses on the image segmentation method based on ACMs.Firstly,we make a review of the image segmentation algorithms.According to the classification of the ACMs,several classical models are introduced in detail.Then,the relative mathematical knowledge of the ACMs is described.Finally,the segmentation algorithm for images with intensity inhomogeneity based on ACMs is studied emphatically.The main contents are as follows:1.A novel active contour model for images with intensity inhomogeneity combining global and local information is proposed.The proposed model makes full use of the regional information of the images,eliminates the influence of inhomogeneity pollution,and uses the correntropy criterion to obtain the local adaptive weight,which effectively improves the image segmentation precision.2.The fast segmentation algorithm based on information geometry is improved.The original fast algorithm used Riemann steepest descent method to change the direction of curve evolution,which greatly improved the convergence rate of the algorithm.However,the original algorithm doesn't take into account the local characteristics of the image,which can not segment the images with intensity inhomogeneity.Therefore,this paper improves the original algorithm and makes full use of the image information to improve the accuracy ofalgorithm segmentation.In the experiment,the segmentation results of different models are quantitatively analyzed by the Jaccard similarity coefficient(JS)and the Dice similarity coefficient(DSC).The better segmentation results,the closer the JS and DSC values to 1.The results show that:1.Compared with the traditional active contour models,the proposed model JS and DSC values are closer to 1,and the number of iterations is not more than 50 times.Relying on the above quantitative and qualitative demonstrations,the proposed model can achieve better segmentation results and has high computational efficiency.In addition,when different initial contours are used,the proposed model can still realize correct segmentation for the inhomogeneity images,whereas the traditional models are easily trapped in the local minimum.These segmentation results demonstrate that the proposed model not only achieves better segmentation effects on the inhomogeneity images,but also overcomes the defects that the traditional model is sensitive to noise and initial contour position.2.The improved fast algorithm using the Riemann steepest descent method to evolve the curve,the number of iterations is no more than 10 times,which indicates that the improved algorithm has a faster convergence rate.In addition,compared with the original fast segmentation algorithm,the improved algorithm combines the image global information and local features,not only can retain the advantages of the original algorithm in the weak boundary image segmentation,but also can produce better segmentation results on the inhomogeneity images.