A new multiresolution classification model based on partitioning of feature space

Multiresolution analysis is a hot topic in the past decade. In this paper, we propose a new multiresolution classification method which adopts a coarse-to-fine strategy both during the training and the testing processes based on decomposing of the feature space. The training algorithm locates the boundary between two classes from coarse to fine by dividing the hypercubes which lie on the boundary step by step. The testing algorithm firstly labels the testing data set by the classifier trained at initial resolution. Then, only those lying on the boundary are labeled at the finer resolution. As an example, an approach named MRSVC is proposed, which exploits support vector machines as the basic classifier. Finally, theoretical analysis and experimental results have substantiated the effectiveness of the proposed method.