利用小波包变换来检测纹理图象上的灰度变化，以得到多分辨率、多方向性的微观统计特征图象，在此基础上用Envelop算法提取基于边缘信息的纹理特征，为了克服纹理特征之间的相关性，采用了子空间分类器对纹理特征图象进行分类，取得了较好的分割效果。

The wavelet packet transform is used to detect the gray variation on textured images, then the multiresolution and multichannel texture features are extracted from every wavelet packets by the Envelop algorithm. These texture images with the same size are classified by the subspace classifier which can overcome the correlation between the feature images . The experiment indicates that the proposed approach provides a good results for textured image segmentation.