Abstract
A method for image classification based on category proportion estimation is proposed. In this method, all pixels in a remotely sensed image are assumed to be mixed pixels (Mixels), and are classified to the most dominant category. Among the E, there exists unconfidential pixels which should be catetorized as unclassified pixels. In order to discriminate them, two types of criteria, Chi squere and AIC, are proposed for fitness test on the pure pixel hypothesis. Experimental result with a simulated dataset show an usefulness of the proposed classification criteion compared to the conventional maximim likelihood criterion and applicability of the fitness tests based on Chi squere and AIC.
