Adaptive Selection Of Image Classi

Recently, the concept of \Multiple Classiier Systems" was proposed as a new approach to the development of high performance image classiication systems. Multiple Classiier Systems can be used to improve classiication accuracy by combining the outputs of classiiers making \uncorrelated" errors. Unfortunately, in real image recognition problems, it may be very diicult to design an ensemble of classiiers that satisses this assumption. In this paper, we propose a diierent approach based on the concept of \adaptive selection" of multiple classiiers in order to select the most appropriate classiier for each input pattern. We point out that adaptive selection does not require the assumption of uncorrelated errors, thus simplifying the choice of classiiers forming a Multiple Classiier System. Reported results on the classiication of remote-sensing images show that adaptive selection can be used to obtain substantial improvements in classiication accuracy.