A new multi-agent system for video objects segmentation and tracking based on spatio-temporal descriptor

The problem of segmentation and tracking is of great importance. It is an important task within the field of computer vision. The goal of this paper is to describe a novel video sequences segmentation and tracking system based on MAS "multi-agent systems" when we use spatio-temporal interest points detection on video frames by HOG3D descriptor to initialize agents. The used agents are supervisor and explorator agents, they communicate between them and they inspire in their behavior from Fast Marching Method.