A Multi-Agents Approach for a Cooperative Image Segmentation

Y. Kabir and Y. Cherfa (Algeria)


computer vision, image segmentation, distributed artificial intelligence (DAI), estimation, multiagents, PVM.


We propose to conceive an image analysis system, based on distributed artificial intelligence (DAI), according to a multi-agents approach. We are especially interested to the first steps system, (image pre-processing and segmentation). we introduce a new segmentation system obtained by the co-operation of a robust edge detector of canny type and a region growing. the cooperation technique allow to exploit the advantages of each method to reduce the drawback effects of the other one. The originality of this method lies in the introduction of an adaptation stage allowing a better choice of the edge detector parameters and the homogeneity criteria thresholds related in a great part to the textured aspect of the image. A multi-agents approach provides an increased flexibility, by separating the various expertises within the system. Each agent is responsible for a precise task and its interactions with other agents are carried out by well defined protocols. Such an implementation makes the system development easier as well as later modification. The developed architecture allow to carry out in parallel a lot of system processes. The system consist of several agents executed as independent processes on several machines connected on a heterogeneous network and communicating through message passing using PVM.

Important Links:

Go Back