How to Complete Any Segmentation Process Interactively via Image Foresting Transform

The segmentation of poorly defined structures in medical imaging and heterogeneous objects in natural images usually call for considerable user assistance. Consequently, automatic results are often far from desirable and interactive repairs become an essential feature to consider. However, how to import automatic results obtained from external processes and complete their segmentation interactively is an issue, since different tools are based on different optimization criteria. Another simpler related problem concerns how to continue a previous segmentation obtained by the same interactive tool. This ability to stop and later resume interactive segmentation sessions is specially important for tridimensional images and video. However, very often crucial data (e.g., the history of user input) are no longer available; or are no longer reliable, as consequence of some post-processing. How to offer a comprehensive recovery and resume capability, comprising all these different scenarios, under the framework of the "Image Foresting Transform" (IFT) is the central focus of this paper.