基于小波变换的多尺度自配准纹理分析

图像分析中多尺度分析与自配准性质与人类视觉相关联，是纹理分析的重要工具。利用小波变换和傅里叶变换对纹理图像进行二次变换域分析构造具有多尺度自配准性质小波系数频谱，然后根据小波系数性质与物理意义对算法进行优化，最后通过小波系数频谱的投影提取图像的纹理特征。利用图像检索与树状结构小波和傅里叶变换方法进行了对比，实验结果表明小波系数频谱具有良好的纹理分析能力。

Texture Analysis Using Multi-scale Auto-registration Based on Wavelet Transform

The multi-scale and the auto-registration are related with human vision,and the important tools especially to texture analysis.The concept about frequency spectrum of wavelet coefficient(FSWC) is introduced which has the capability of auto-registration based on multi-scale.Then it is optimized according the characteristic of wavelet coefficient.The last step is to extract the feature vectors from the FSWC.The algorithm is tested by image retrieval compared to tree-structured wavelet and Fourier transform.The experiment result shows that FSWC is a promising method to texture analysis.