ABSTRACT

Deformable models (DM), or active contours and surfaces, which are powerful tools of image segmentation, may incorporate various stochastic models of visual appearances and shapes of goal objects. Parametric or nonparametric (geometric) DMs are evolving towards intricate object boundaries under the external and internal forces outlined in brief in Chapter 2 (Section 2.4). If external forces depend on joint probabilities of image signals and region labels in individual points within and outside the object, such an active boundary could be called a stochastic DM.