摘要
提出了一种针对不容易描述的不规则特征的提取方法:采用贝叶斯启发式学习方法提取图像的聚类变量和等价变量作为特征;用网格划分技术过滤和释放位于稠密超方格的数据项,从而有效减少内存需求、大幅度降低计算复杂度。将此方法应用于医学图像分类器中的特征提取部分,实验结果表明大大地提高了分类的准确率。
A method of extraction of irregular features which are difficuh descriptive is presented.Bayesian networking learning is used to extract the clustering variances and equivalent variances as the irregular features for images;Gridpartition approach is discuss to filter out and release the data objects in the crowded grids,which leads to less memory space and simplifies computational complexity.A medical image classifier is designed and makes use of this method to extract irregular features.Its result shows that irregular features improve the accurate rate of image classifier.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第28期52-54,96,共4页
Computer Engineering and Applications
基金
江苏大学科研基金资助(编号:04KJD001)
关键词
图像挖掘
不规则特征提取
信念网
网格划分
image mining,irregular feature extraction,Believe Networking,Grid-partition