Image segmentation based on ensemble learning

Image segmentation is an important tool in image processing and can serve as an efficient front end to sophisticated algorithms and thereby simplify subsequent processing. This paper proposes a new method based on ensemble learning. In this method, to eliminate the useless and redundant features, the simulated annealing algorithm is used to select the important features as classifiers. And training set is also selected by a cluster algorithm not only to improve the accuracy of classification but also to reduce the structure of final decision tree. With the important features and representative training set, the decision tree is built. The criterion of this tree is selecting the property which has smallest error rate. Finally, an experiment with aerial image is implemented. And the results demonstrate that our method shows a higher accuracy of segmentation.