基于颜色特征的自适应图像分类算法及其应用  

Color Feature-based Adaptive Classification Algorithm and Its Application

作　　者：余雄[1] 张著洪[1] 

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025

出　　处：《贵州大学学报（自然科学版）》2017年第1期62-65,共4页Journal of Guizhou University:Natural Sciences

基　　金：国家自然科学基金项目资助(61563009)

摘　　要：针对基于内容的图像分类问题,探讨自适应图像分类算法。引入最大类间方差法去除图像背景噪声;依据惯性比和金字塔结构模型,获得目标特征值向量模型,进而借助K-means算法思想和目标特征向量获得自适应图像分类算法。该算法具有图像分类数自适应确定、分类效果不依赖于初始中心的优点。比较数值结果表明,此算法分类效果较好且无参数依赖性。This work investigates an adaptive image classificationalgorithm for the problem of content-based image classification. In this algorithm,the Otsu method was adopted to remove background noises,while a target feature vector model was designed based on a Pyramid structure model and the version of inertia ratio. After those,one such algorithm,with the merits of adaptive imageclassification and no sensitivity to initial classification centers,was obtained by the idea of the K-means algorithm and targetfeature vectors. Comparative experiments have validated that it can well perform image classification without parameter dependence.

关 键 词：最大类间方差法 HSV空间 惯性比 K均值算法 图像分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]