Study On Active Contour Models Integrating Local And Global Information

As a subject which is involved broadly in many fields,image processing has received much attention of scholars. Image segmentation is an important subject in the field of image processing.The more accurate segmentation results can bring the better analysis. Up to now,wealth of models and algorithms has been designed to handle the segmentation problems of a variety of images.In the field of image segmentation,active contour models based on level set method have been testified to be one class of efficient methods,with the merit of the efficiency of numerical calculation and simplicity,the active contour models have been widely studied in the past two decades.Active contour models can be divided into two categories:edge-based models and region-based models.The region-based models can segment the noisy images and are robust to the initial contour.This paper firstly described some classical active contour models(such as LIF model,and GIF model),then we proposed a model to segment images with intensity inhomogeneity and complicated background.Our main work is as follows:By combining the global information and local information,we construct a new active contour model,which contains a local term,a global term and a regularization term.The global term can make the model robust to the initial contour.The local term is based on the difference image between Gaussian convolution image and the original image, which has the advantage of handling the images with intensity inhomogeneity.The regularization term is used to smooth the level set function and avoid the re-initialization procedure.Experiments on synthetic and real images illustrate that the proposed model can well segment images with intensity inhomogeneity,and is robust to the initial contour.