Variational methods for image segmentation by using the level set function and its applications

Abstract

This dissertation concerns the variational methods for image segmentation by using the level set function. We classify images into two categories: two-phase images and multi-phase images. The multi-phase segmentation problem is unstable in the sense that the segmentation result depends significantly on the number of different phases given a priori. This is a critical issue in the multi-phase segmentation and we want to automatically obtain the phase number through the process. First, we survey the previous two and multi-phase segmentation models and then propose a new segmentation method for multi-phase images, which can determine the phase number in the process. Finally, we show some applications of a level set based segmentation method in realistic problems: the automatic number plate recognition and the medical field.