Image segmentation with shape priors: Explicit versus implicit representations

Image segmentation is among the most studied problems in image understanding and computer vision. The goal of image segmentation is to partition the image plane into a set of meaningful regions. Here meaningful typically refers to a semantic partitioning where the computed regions correspond to individual objects in the observed scene. Unfortunately, generic purely low-level segmentation algorithms often do not provide the desired segmentation results, because the traditional lowlevel assumptions like intensity or texture homogeneity and strong edge contrast are not sufficient to separate objects in a scene. To overcome these limitations, researchers have proposed to impose prior knowledge into low-level segmentation methods. In the following, we will review methods which allow to impose knowledge about the shape of objects of interest into segmentation processes.