ABSTRACT
This chapter focuses on the extraction of a single object. It shows how the image graph may be constructed to accommodate general target specification and reviews several of the most prominent models for targeted image segmentation, with an emphasis on their commonalities. The chapter shows how the various types of target specification may be used to identify the target object. It focuses on the specification of target information to define the segmentation of a single object. The chapter reviews a variety of ways to provide a target specification that have appeared in the literature. It assumes that the edge weights represent affinity weights. The chapter considers method that adds an explicit shape bias by using training data or a parameterized shape. It also assumes that the training data contains both images and segmentations of the desired object. The chapter examines the relational target specification and so slightly expands the scope beyond singleobject segmentation.
