engleski

Autonomy Oriented Computing Applied in Color-based Image Segmentation

This paper presents an Autonomy Oriented Computing (AOC) approach to color image segmentation. Basic elements in AOC systems are an environment and autonomous entities. Environment, in our case, consists of the three layer 2D lattice which contains color images in RGB and HSI color spaces as well as storage cells through which entities interchange information. Environment serves as the place where autonomous entities roam and operate. Entities exhibit a number of reactive behaviors such as diffusing, self-reproduction, waiting or dying. Goal of each entity is to locate and label image pixels whose color satisfies skin-like criteria. Entities which are successful in locating such pixels have an opportunity to produce offspring entities and place them within their neighborhood, where they are most likely to find other skin color-like pixels. The entities that don’ t succeed in locating any skin color like pixel vanish from the system when their ages exceed their lifespan. Results of all entity families are combined into regions. Experiments, based on a program simulator, were run over set of frontal face images from the XM2VTS database.

Autonomy Oriented Computing (AOC); Color-based Image Segmentation; Autonomous entities

nije evidentirano