On the use of social agents for image segmentation

In the literature, there are a lot of methods for image segmentation. Unfortunatelly, they are often limited in their capacity to treat image odtained by an acquisition system (Optical, X-Ray, IRM, . . .). Thus, many of them are dedicated to particular solutions and there is no generic method for solving the image segmentation problem. In this paper, we present a way to implement segmentation methods which use models coming from biology: social spiders and social ants which are implemented by a multi-agent system. After a presentation of the principles of these two methods, we will quickly present two another ones: Region Growing and Otsu thresholding methods, in the aim to compare their results. The simulation of these methods shows results that are promising. Some perspectives have been retained in order to overcome agent-based methods for having a robust segmentation technique.