An unsupervised model for image classification

In this paper an unsupervised classification model based on level set method is presented. In recent years many classification algorithms based on level set method have been proposed for image classification. However, all of them have defects to some degree, such as parameters estimation and re-initialization of level set functions. To solve this problem, a new model including parameters estimation capability is proposed. Even for noise images the parameters needn't to be predefined. This model also includes a new term that forces the level set function to be close to a signed distance function. Therefore it saves the time for classification. The proposed model has been applied to both synthetic and real images with promising results.