A new image segmentation method based on genetic algorithm of multilayer perception

To our best knowledge, classical image segmentation methods are serial in essence, with poor robustness and limited application field. Genetic image segmentation algorithm was proposed in recent years, which can overcome those drawbacks in certain extent, but with single resolution, regions with different extent of details can not achieve good segmentation results simultaneously. We propose in this paper a new image segmentation method to solve this problem. It is based on multiresolution concept with multilayer structure. On each layer, genetic method is used to achieve segmentation result. Then from coarse to fine, the final result can be obtained. The whole process is proved to be robust and parallel. Compared with genetic segmentation method with single resolution, it can achieve better results.