Multistage Decision Level Image Fusion Technique for WorldView-2 Images

Abstract

This research proposes a decision level fusion scheme to integrate the classifiers which use different depiction and these depictions can be considered as special cases to combine them to make a decision. A solitary classifier cannot deal with the wide variety and scalability of information in any application area, thus there is a need for integration of classifiers. Most recent classification methods use the integration of classifiers and fuse the decisions provided by the individual classifiers, by using the relevant feature set for the task. The proposed decision level Naive Bayes fusion (DFNB) considers the both supervised and unsupervised classification techniques. Two levels of fusion are done here, a first level decision fusion of a pair of classifiers is made by using the weighted majority voting method (WMV). And a second level decision fusion of a set of classifiers is done by using the Navia Bayesian fusion method. The results of the first level decision fusion of a pair of classifiers are compared with the second level decision fusion at the end.