关键词: 图像分类, JointBoost, 图像到类距离, 原型特征集

Abstract: Image classification on the basis of image-to-class (I2C) distance metric is a novel method． However，its classification performance needs to be further improved． In this paper，a new image classification method on the basis of JointBoost I2C distance metric is proposed． In this method，a prototype feature set with representative sam-ples is generated,which makes the calculation of distance from the test image to the set more effective． Then，on the basis of JointBoost algorithm，multiple I2C distance metrics are combined to generate a strong classifier for in-tegrating spatial information． Experimental results show that the proposed method is of higher performance for image classification．

Key words: image classification, JointBoost, image to class distance, prototype feature set