Research On Image Segmentation Based On Variational Method

Images segmentation is one of the key issues in computer vision, its scope of application is very broad, and it appears in almost all areas of the image processing. Traditional non-model methods are difficult to meet the applications of complicated image segmentation, for example, medical image segmentation, because of their incomplete border region segmentation, and the lack of a priori knowledge. So, images segmentation based on variational method is emerging.Active Contour models, Mumford-Shan model and Chen-Vese model are Traditional segmentation models based on variational models. The traditional variational model obtained great success in dealing with the low level visual information (grey or color information) images. Therefore, in recent years, research of segmentation is focused on the traditional model integrated with the higher visual information. Segmentation models based on shape priori knowledge add the shape priori knowledge to the traditional model, the objects which are blocked or loss of information will be correctly segmented.Texture difference is the difference between the texture on objects and background of the image. From the frequency analysis point of view, the difference lies in the frequency, scale and direction. Extracting texture information from images used texture structure tensor or Gabor filters. The feature space composed of extracted texture information, the feature space needs pretreatment to remove noise. The variational method based on texture information is one that integrates the texture information into traditional model.