Research On Image Segmentation Method Combining Fractional Differential And Active Contour Model

Image segmentation is the process of segmenting a given image into different regions according to its characteristics,and extracting the target of interest from the background.In the field of image segmentation,image segmentation based on Geometric Active Contour model has been one of the most active research hotspots.This paper mainly studies two Geometric Active Contour based on local regions,especially the problems existing in the segmentation of weak edges,weak texture regions and complex structure images,and related improved algorithms are proposed.Detailed research work is described as follows:1.Local Image Fitting(LIF)model uses local image fitting energy to obtain local information,which drives the evolution curve to move,lacking global characteristics.Therefore,when the model divides images with weak edges and weak texture regions,the requirements for initial contour curve selection are higher,and the robustness to noise is poor.The Weight Global Image Fitting(WGIF)model obtains energy functionals by introducing the difference between the weighted global fitted image and the original image,which lacks local information.Aiming at the problems of LIF model,an image segmentation algorithm is proposed,which based on fractional differential WGIF model and LIF model.The fractional differential gradient fitting force of the global image information of WGIF model and the local gradient fitting force of LIF model constitute the driving force of the evolution curve in the improved method.The driving force of the evolution curve is added,which drives the evolution curve to move towards the target contour continuously.Through theoretical analysis and experimental results,the proposed model can effectively segment the weak edge and weak texture images,and the segmentation accuracy and efficiency are also improved,and the requirements for the initial contour selection of the evolution curve are greatly reduced,and also improve the noise robustness.2.Local Gaussian Distribution Fitting(LGDF)model describes the gray level distribution of local image by means of Gauss distribution statistics of mean and variance.Compared with the region scalable fitting(RSF)model and LIF model,this model can achieve better segmentation results for gray scale inhomogeneous images,noise images and relatively complex images.However,the model lacks globalinformation and is sensitive to the selection of initial contour curve.Especially in the case of segmentation of images with weak edges and texture regions,the local optimal solution may appear,and the robustness to noise is poor,which leads to the failure of segmentation.In order to improve the segmentation performance of LGDF model,the global Grümwald-Letnikov(G-L)fractional gradient fitting term is fused in LGDF model with local fitting term,which can enhance the gradient information of weak edge and texture regions,and thus enhance the robustness to initial contour curve and noise.The coefficients of global and local terms are determined by adaptive weight function,which improves the efficiency and accuracy of segmentation for uneven gray images.An adaptive fractional order mathematical model is constructed by using image contrast,information entropy and gradient modulus to improve segmentation efficiency.The new model enhances the robustness to the initial contour curve and noise,and improves the segmentation accuracy and efficiency for gray scale inhomogeneous images.