Interactive foreground extraction using graph cut

Note, this is an extended version of chapter 7 from the book: Markov Random Fields for Vision and Image Processing, MIT Press [6]. In this Technical Report, references to other chapters are with respect to the book. The differences are, a new section 4.3 and extra details in section 3.2 and 3.3 The topic of interactive image segmentation has received considerable attention in the computer vision community in the last decade. Today, this topic is very mature and commercial products exist which feature advanced research solutions. This means that interactive image segmentation is today probably one of the most used computer vision technologies worldwide. In this chapter we review one class of interactive segmen-tation techniques, which use discrete optimization and a regional selection interface. We begin the chapter by explaining the seminal work of Boykov and Jolly [9]. After that the GrabCut technique [36] is introduced, which improves on [9]. GrabCut is the underlying algorithm for the Background Removal tool in the Microsoft Office 2010 product. In the third part of the chapter many interesting features and details are explained which are part of the product. In this process several recent research articles are reviewed. Finally, the Background Removal tool, as well as [9, 36], are evaluated in different ways on publicly available databases. This includes static and dynamic user inputs. 1