基于Radon变换和EMD的旋转不变纹理分类"

A Rotation Invariant Texture Classification Based on Radon Transform and EMD

提出一种结合Radon变换和经验模态分解(EMD)的旋转不变纹理分类方法。对纹理图像进行Radon变换,使用EMD分解成一系列固有模态函数(IMF),用修改过的能量公式计算各IMF的能量并组成特征向量,利用k-近邻对纹理图像进行分类。通过对Brodatz纹理库中的图像进行分类实验,表明所提方法能够有效地进行旋转不变的纹理分类,较之单一的Radon变换分类有了显著的提高。

A rotation invariant texture classification method using Radon transform and Empirical Mode Decomposition （EMD） is proposed. First, texture image is translated by Radon transform to convert the rotation to translation. Secondly, the transform result is divided into several Intrinsic Mode Functions （IMF）. Then the invariant feature is composed of the energies of each IMF using a modified energy equation. At last a k-Nearest Neighbor Distance is applied to classify the rotation textures. The result shows the proposed method can greatly improve the correction rate of rotation texture classification after applying this method on Brodatz.