Local And Global Information Adaptive Fusion Active Contour Image Segmentation

Image segmentation is the basis of image processing.It is a process to divide the image area with some unique features from the whole area.A geometric active contour model is calculated based on the image low-level features and basic mathematical theory,so it is a hot spot in the image segmentation.And geometry active contour model based on the region has become the focus of research because of its fast convergence speed,high segmentation efficiency and good robustness in recent years.Because the CV model cannot process the image with in-homogeneous gray scale,the LBF model is too sensitive to the initial contour,the weight selection of the LGIF model is too complex and so on,this paper presents the adaptive fusion active contour model based on local and Global Information(Local and Global Information Adaptive Fusion Active Contour Model),Called LGAIF model,introduces the different means and variances to the CV model and the LBF model respectively and add the energy norm function of the two models together to form a LGAIF model.Firstly,the paper introduces the adaptive fusion coefficient to optimize and improve the LGAIF model performance.The proportion weight of the local information to the global information is adjusted dynamically according to the gray-scale heterogeneity of the image and speeds up the evolution to segment the target area accurately.Secondly,the variable-step algorithm is introduced to the LGAIF model.In iterative methods,to adjust dynamically time step overcomes the problems such as the evolution low speed and shock the fixed length brought.In addition,a new punishment is introduced to improve the segmentation efficiency without initializing the level set function.Finally,this paper compares the LGAIF model presented in the paper with the CV model,the LBF model,the LGIF model,the LGIF model and the previous LGIF model and the segmentation result is analyzed objectively and quantitatively in term of the segmentation time,the iterative number and the similarity.The final result shows that the LGAIF model presented in the paper not only has the better robustness in selecting the position of the initial contour curve,but improve the segmentation accuracy and the efficiency.