基于Contourlet变换和不变矩的纹理特征提取与分类

Texture Feature Extraction and Classification on Contourlet Transform and Invariant Moments

本文提出了一种基于contourlet变换和不变矩的特征提取方法.它首先对图像进行contourlet变换用以多尺度多方向分析,然后提取变换后各子带系数的统计特性和不变矩特征,以构造特征矢量,在此基础上,根据不同子带特征分类能力的不同,对各子带数据的离散程度进行加权处理,为分类能力强的特征量赋予较大的权值,得到新的特征向量,最后利用RBF神经网络作为分类器进行分类.实验结果证明了该方法的有效性和良好的分类能力.

This paper proposed a feature extraction method based on the contourlet transform and Invariant moments.First it should be transformed using contourlet transform for analysis with multi scale and multi directional,and then extract the statistical characteristics and moment invariant features of the subband coefficient,constructed as feature vector.The feature vector is weighted according to the degrees of classificion,and the feature with higher classification ability has bigger weight,which are calculated to get some new feature vectors.At last,classify the extracted feature vectors by RBF Neural network which works as a classifier.The experimental results proved the effectiveness of the methods and the better classification ability.