Active Nets and Cubes

Active models have attracted a considerable amount of interest in recent years. Many kinds of active contours and surfaces as well as energy minimization schemes have been presented. This paper describes two extensions of this family of active models called active nets [11] and active cubes. These models are 2D nets and 3D cubes of nodes controlled by physical models in the tradition of the SNAKE [6] family of active models. The advantage of 2D nets is their ability to split-up in a well-behaved manner. This allows one net to extract two or more objects from a scene. This paper presents a robust algorithm capable of controlling this splitting. It also shows how 3D cubes can be used both to extract 3D object shape and as the basic data-structure for simulated medical operations. An improvement of the Greedy energy-minimization algorithm [12] inspired by Simulated Annealing is also presented.