In this paper we present a general framework for object detectionand segmentation. Using a bottom-up unsupervised merging algorithm,a region-based hierarchy that represents the image at differentresolution levels is created. Next, top-down, object class knowledgeis used to select and combine regions from the hierarchy, in orderto define the exact object shape. We illustrate the usefulness of theapproach with four different object classes: sky, caption text, trafficsigns and faces.