【摘要】 针对图像分类中使用视觉词袋直方图进行分类时忽略图像颜色信息缺点,该文提出在一类图像的HSI彩色空间上,通过H分量和S分量构建二维最大熵模型,并将得到的二维最大熵分布作为该类样本的底层参考特征向量,从而将待分类的图像运用欧式准则与底层特征向量进行匹配,最终实现图像分类算法.实验表明,该文所提分类算法比基于视觉词袋直方图分类算法具有更高的查准率.更多还原

【Abstract】 A 2-D maximum entropy model is proposed to overcome the weakness of BOVM(bag of visual words)histogram which always neglects the information of image color.The 2-D maximum entropy mode of a class of image is built by H component and S component in HSI color space,and what’s more,the corresponding 2-D maximum entropy distribution is the bottom reference feature vectors,which is used to match with an input image by Euclidian criterion in image classification algorithm.Experiments illustrate that the algorithm presented in this paper has a higher image precision than the classification algorithm based on BOVM.更多还原