Probabilistic interactive segmentation using multiphase level sets

This paper presents methods to incorporate user input into a probabilistic multiphase set segmentation framework. The first method allows the user to input preferred labels to regions of the image. The second method allows the user to input boundary segments between two regions. This paper improves on existing methods by allowing an arbitrary number of phases and by formulating a probability model that does not assume the image and user input are independent. The methods presented are applicable to an image of any dimension with any number of phases.