二元树复小波在图像特征提取和分类中的设计与应用

Design and Application of Image Feature Extraction and Classification by Using Dual-Tree Complex Wavelet Transform

通过二元树复小波变换对图像进行4尺度分解,提取每一尺度下代表6个方向的高频带子图小波系数模的均值和标准方差组成48维的特征向量,利用支持向量机的一对一多分类算法对Brodatz图像库中的112幅图像进行了纹理特征提取和分类实验,结果表明二元树复小波变换提取的图像特征能有效提高图像的分类精度.

The dual - tree complex wavelet transform was used to decompose images with four scales, and the 48 - dimensional feature vector was generated by computing mean and standard deviation from wavelet coefficients of six - direction high - frequency subbands of each scale, using one - vs - one algorithm of the support vector machine, the texture feature extraction and classification experiments were done by using 112 images of Brodatz image database. The results show that the image features of the dual - tree complex wavelet transform can effectively improve the image classification accuracy.