A biologically motivated paradigm for scene segmentation

In this paper, we propose a new paradigm to be applied to solve the image segmentation problem. This paradigm is neurophysiological based and exploits the physical properties of a network of dynamically coupled chaotic oscillators to accomplish scene segmentation. Thus, an object in a scene is associated with time evolutions of synchronized chaotic oscillators, while different objects are discriminated by different sets of synchronized chaotic oscillators that are desynchronized with one another. Computer simulations that support our paradigm are presented.