将四元数小波变换(QWT)和多分形相结合进行纹理分类,充分利用了QWT的旋转不变特性和纹理图像的多分形特性,能弥补传统的应用小波变换进行纹理分类时缺乏将输入图像分解成多个方向的不足.通过对UIUC数据库中的纹理图像分类,表明四元数小波与多分形相结合的方法具有较高的分类精度,平均分类正确率可达96.69％,是一种合理有效的纹理分类方法.%The paper incorporated the multifractal analysis method into the idea of Quaternion Wavelet Transform ( QWT), which took advantage of the rotation-invariant properties and multifractal properties of texture image, and could make up for the lacks of ability to decompose input image into multiple orientation in texture classification when using wavelet transform. The experiment of texture classification using the images from UIUC shows the method has higher classification accuracy and the average correct classification rate is 96. 69%. It proves this texture classification method is reasonable and effective.
展开▼
