多进制Haar与斜Haar小波在纹理分类中的应用

Texture classification based on M-mand Haar wavelet transform and slant Haar wavelet filter banks

基于Walsh变换和斜变换给出了一种构造多进制Haar小波和多进制斜Haar小波的方法,并利用这些变换特有的快速算法加快了小波分解速度.实验中用该类小波分解图像,然后提取纹理图像的特征量,使用最小距离分类器进行纹理分类并比较了构造出来的各类小波的分类正确率,给出了选择小波基的原则.实验结果表明多进制斜Haar小波滤波器组具有更佳的分类效果.

In this paper,a kind of M-Band wavelet based on walsh transform and slant transform is introduced. The wavelet filter banks based on Walsh transform and slant transform are not only constructed easily and selected flexibly, but also can accelerate wavelet transform based on the fast algorithm of these transforms. It can improve the practicability of M-band wavelet. In the numerical test, we decompose texture images into some parts with these two wavelet filter banks, then extract texture features and classify texture based on the minimum distance classifier. The principle of the wavelet selected can be concluded. The numerical test also show that the slant Haar filter banks can give a satisfactory result.