Automated parameterization of active contours: A brief survey

Abstract

Active contours yield segmentation results which depend on an initial empirical parameterization stage. The latter is a tedious and time-consuming process that requires technical skills from the end user. Automated adjustment of active contour parameters is still a challenging issue. This survey reviews state-of-the-art active contours which attempt to cope with the issue of empirical parameterization, so as to secure the objectivity and robustness of the segmentation results. Numerous attempts utilize information associated with contour evolution and shape priors, whereas others are hybrid, driven by both empirical and automatically obtained parameter settings. Most recent models are spatially varying and versatile regarding the application and the energy functional to be minimized.