Adaptive approach to image segmentation based on the region features

A new adaptive image segmentation method was put forward which combines both the image stochastic hypothesis model and the conditional entropy of system to update adaptively the weight of self-organizing feature-mapping-network on the basis of the stochastic characteristics of image, hence the problems (such as: the parameters need to be adjusted artificially; it's quite difficult to obtain the optimal solution) in the conventional methods can be then solved. The experimental results show that the method proposed has the advantages of both robustness and effectiveness for the image segmentation of varieties of complex images.