基于非监督纹理分割下的图像查询  

An Unsupervised Segmentation Framework for Texture Image Queries

作　　者：杨利峰[1] 胡茂林[2] 

机构地区：[1]安徽大学智能计算与信号处理实验室,安徽合肥230039 [2]安徽大学数学与计算科学学院,安徽合肥230039

出　　处：《计算机技术与发展》2006年第5期77-79,共3页Computer Technology and Development

摘　　要：随着图像数据库的广泛应用,基于图像纹理的查询已经变成了一项重要也是复杂的工作。为了更简便地查询图像,文中基于图像纹理的非监督分割,提出了新的图像分类方法。通过对每一纹理图像进行这种非监督纹理分割,提取纹理的特征向量,然后基于这些参数,运用假设检验进行纹理的分类和融合。通过对MT Vistex和Brodatz数据库中选取的纹理图像进行分类测试,证明该方法是十分有效的。Texture image query is becoming an important and complex task. In this paper, a novel unsupervised segmentation for texture image queries is presented, By applying the unsupervised texture segmentation method on each texture image,a set of texture feature pa- rameter for texture image can be extracted automatically. Based upon these parameter,an effective multi- filter query strategy which al- lows the user to issue texture - based image queries is developed. The test results of the proposed framework on texture image obtained from the MT Vistex and Brodatz database are present to show its effectiveness.

关 键 词：非监督纹理分割 特征参数 图像查询 多重滤波器 

分 类 号：TN911.73[电子电信—通信与信息系统]