摘要
针对频域变换的旋转不变纹理特征,提出Radon变换结合树型小波变换的纹理特征提取方法。以Radon变换处理影像将影像纹理的旋转转化成平移信息,对变换后的影像进行非完全树型小波变换。此时可以将能量作为旋转不变纹理特征,但得到的小波系数所携带信息较少,所以本文采用小波系数直方图作为纹理特征。采用此方法对居民地进行提取,实验结果表明利用该方法能得到比较有效稳定的旋转不变纹理特征,并且对不同的中小比例尺遥感影像都有较好的适应性。
Considering the invariant rotation texture features of frequency-transformation,a texture features＇ extraction algorithm of Radon transformation based on tree-structured wavelet transformation is proposed.Images are processed by Radon transformation from the rotation of image texture to shifting information.Then the processed images are transformed by incomplete tree-structured wavelet.Now we can consider the energy as invariant rotation texture features,but the information achieved of wavelet transforming coefficient is too less.So this paper adopts wavelet transform coefficient histogram as texture features.Using this method to extract the residential area,experiments show that we can achieve more efficient and steady invariant rotation texture features,and this method has good adaptability for different medium and small scale remote sensing images.
出处
《测绘与空间地理信息》
2012年第1期150-153,共4页
Geomatics & Spatial Information Technology
关键词
纹理特征
频域变换
RADON变换
树型小波变换
texture features
frequency-transformation
Radon transformation
tree-structured wavelet transformation
