摘要
提出一种基于多示例学习的图像表示方法,将图像作为多示例包,用高斯滤波器将图像滤波并取样为由颜色区域构成的矩阵,使用单颜色及相邻区域(single blob with neighbors)的包生成方法。根据用户选择的实例图像生成正包和负包,使用MIL-SVDD_I和MIL-SVDD_B算法进行实验。实验表明该图像表示方法是可行的。
In this paper,a multi-instance learning-based Image representation approach is presented.The whole image is regard-ed as a multi-instance bag with the Gaussian filter to color image filtering and sampling the regional composition of thematrix of the ground.Single blob with neighbors hypothesis class to describe the image.Images posed by the user aretransformed into corresponding positive and negative bags and the MIL-SVDD_I and MIL-SVDD_B algorithm are employedfor image classification.The experimental results are promising.
出处
《工业控制计算机》
2012年第6期74-75,共2页
Industrial Control Computer
关键词
基于内容的图像检索
多示例学习
图像表示
图像分类
content-based image retrieval,multiple-instance learning,image representation,feature extraction,image classification
