RELAXATION METHODS FOR SUPERVISED IMAGE SEGMENTATION

We propose two methods for supervised image segmentation: supervised r elaxation labeling and watershed-driven relaxation labeling. The metho ds are particularly well suited to problems in 3D medical image analys is, where the images are large, the regions are topologically complex, and the tolerance of errors is low. Each method uses predefined cues for supervision. The cues can be defined interactively or automaticall y, depending on the application. The cues provide statistical region i nformation and region topological constraints. Supervised relaxation l abeling exhibits strong noise resilience. Watershed-driven relaxation labeling combines the strengths of watershed analysis and supervised r elaxation labeling to give a computationally efficient noise-resistant method. Extensive results for 2D and 3D images illustrate the effecti veness of the methods.