Supervised image classification based on AdaBoost with contextual weak classifiers

AdaBoost, one of machine learning techniques, is employed for supervised classification of land-cover categories of geostatistical data. We introduce contextual classifiers based on neighboring pixels. First, posterior probabilities are calculated at all pixels. Then, averages of the posteriors in various neighborhoods are calculated, and the averages are used as contextual classifiers. Weights for the classifiers can be determined by minimizing the empirical risk with multiclass. Finally, a linear combination of classifier is obtained. The proposed method is applied to artificial multispectral images and shows an excellent performance similar to the MRF-based classifier with much less computation time.