摘要
针对任何单一特征都不能完整地表示医学图像的信息,提取医学图像的颜色特征、纹理特征以及区域形状特征,更多地保留图像的各种信息。并对提取的特征利用主成分分析(PCA)方法进行特征级的数据融合。针对维数对PCA的计算量影响,在PCA融合之前利用模糊方法进行特征的粗选择,有效地降低了特征维数。以肝脏B超图像为研究对象进行实验,结果表明,融合后的数据维数有极大的改善,识别效果良好。
Since any feature alone cannot wholly express clinical image information,the thesis extracts clinical image＇s color features,texture features and region shape features in order to further retain various image information. Then it utilizes principle component analysis（ PCA） method to carry out feature-level data fusion with the extracted features. Taking into consideration the influence of dimension on PCA computational volume,before PCA fusion it utilizes fuzzy method to carry out coarse selection for the features,so that the feature dimension number is efficiently decreased. Taking liver B ultrasound images as objects to study,the thesis carries out experiments. The results show that there are significant improvements over the fused data dimension number,and the recognition effects are good.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第4期239-243,共5页
Computer Applications and Software
关键词
特征提取
特征选择
PCA特征级融合
医学图像识别
Feature extraction Feature selection PCA feature-level fusion Clinical image recognition
