Research On Active Contour Segmentation Model Combining Local And Global Information

Image segmentation is that extracts regions of interest or contours from the objected images,is the vital step in image identification and image understanding,as well as an important research direction in image processing field.So far,there have been thousands of image segmentation algorithms proposed,and the active contour models based on variational level set method have shown more advantageous,in which the active contour model based on region,is one of the hot issues.The region information of images and the variational level set method theory are employed by region-based model,to transform the segmented problem into a partial differential equation problem about level set function.Although the methods have many successful applications,there are still some problems need to be solved,such as the segmented results dependence on the selection of the initial contours,the unsatisfactory segmented results for in-homogeneous images(non-uniform gray-scale images).In addition,the global information of images can increase the robustness to the noise and reduce the dependence on the initial contours,and the local information of images can improve the segmented ability for the in-homogeneous images.Thus,the segmented models are built by combining the local information of images with the global information of images,which is more advantageous than the local information of images or the global information of images is utilized singly.The contributions of this paper are as follows:1.An improved Local Image Fitting(LIF)segmentation model is proposed by introducing the global information of images.The LIF model only considers the local information of the images,which results in the segmented results highly dependence on the selection of the initial contours.Aiming at this problem,an improved segmented model is proposed by combining a simplified global data fitting items with a local data fitting item.Finally,experimental results show that the proposed model can reduce the dependence on the selection of the initial contours and increase the robustness to noise.2.An new active contour segmentation model is proposed by improving signed pressurefore(SPF)function.The fractal order differential information in Fourier domain is introduced as the local information and then the SPF is built by combining the global information.It varies in-1 and 1,which can effectively control the expansion and contraction of the evolution curves.Due to the fractional order differential information has the property of preserving and enhancing the low frequency information,the proposed segmented model will be good for segmenting in-homogeneous images.Finally,experimental results show that the proposed model performs better than other models in both synthetic and real images.3.An adaptive weighting active contour segmentation model is proposed by combining local and global energy fitting.The difference image's energy fitting is employed as local term in the proposed model.The difference image is constructed by the fractal order gradient image,which can enlarge the difference between the foreground of images and the background of images.At the same time,the famous Chan-Vese model's energy fitting is employed as global item of the new model,and an adaptive weighting parameter is introduced to adjust the role of local item and global item.Finally,experimental results show that the proposed model can obtain more ideal segmented results,and spend less times.