肝癌超声图像识别的特征提取  

Extracting Features for Ultrasonic Liver Cancer Images

作　　者：陈菲[1] 

机构地区：[1]四川绵阳西南科技大学计算机科学与技术学院,621010

出　　处：《微计算机信息》2006年第10X期272-274,101,共4页Control & Automation

基　　金：国家级火炬计划项目(2004EB011224)

摘　　要：本文提出了识别肝癌超声图像的一种特征提取。它将共生矩阵和多分辨率提取分形特征方法结合,提取空间灰度独立矩阵、灰度差分统计、Laws纹理能量度量、傅立叶能量谱等特征来实现肝癌超声图像的识别。实验证明,这种有效的特征提取对超声正常肝和肝癌图像具有较高的分类正确率,在对缺乏病理专家的医院和远程医疗信息系统中的应用提供了坚实的理论基础。A study about classification of ultrasonic live cancer images is proposed by the texture features, which combines the cooccurrence matrices with Muhiresolution fractal feature .Many features have extracted including the spatial gray_level dependence matrices ,the gray-level difference statistics, Laws＇ s texture energy measures and the Fourier power spectrum .we have completed to classify normal liver image and ultrasonic cancer live images. The experiment results also show that this study is very effective to decide. it is sound basis to the hospital absence of professional and the telemedicine information system.

关 键 词：感兴趣区域 Bayesian决策 图像识别 纹理特征提取 肝癌超声图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]