AdaBoost Learning Based-on Sharing Features and Genetic Algorithm for Image Annotation

Image classification approach is one promising technique used for image annotation. In order to improve image annotation accuracy, recent researchers propose to use AdaBoost algorithm for the ensemble of classifiers and the weak classifiers in it are constructed on sharing features associated with class subsets. We use all the 25 image low-level features of multimedia content description interface to present images. Genetic algorithm is used to select optimal sharing features. As the exhaustive search of all the possible subsets results in expensive computation cost, a variant of best-first approach is used to reduce search space. AdaBoost.M1 algorithm is used to generate the ensemble classifier and k-nearest neighbor classifier is used as base classifier. The results of experiment over 2000 classified Corel images show that the algorithm has higher annotation accuracy.