|Description:||Given two images of the same scene but from different modalities, the task is to find their common segmentation. This problem is encountered for example in multimodality imaging, or as a subtask of iterative segmentation/registration. One possible formulation is that we want to approximate as well ass possible the target image using a segmentation and an appearance model for each class. We shall study various variations of this idea, e.g. how the approximation quality should be measured and how the resulting optimization should be performed.