Wu Ke, Garnier Carole, Coatrieux Jean-Louis, Shu Huazhong
Laboratory of Image Science and Technology, Department of Computer Science and Engineering, Southeast University, China.
Annu Int Conf IEEE Eng Med Biol Soc. 2010;2010:5581-4. doi: 10.1109/IEMBS.2010.5626789.
A moment-based approach is proposed for texture analysis of medical images. The neighborhood of texture pixel is calculated by moments for texture feature extraction. After verified on Brodatz textures, the moment-based texture analysis method is applied on CT liver scan classification and prostate ultrasound segmentation. A support vector machine and a multi-channel active contour model are used in our application. The result shows that our method works well but still have some limitations.
提出了一种基于矩的医学图像纹理分析方法。通过矩计算纹理像素的邻域以进行纹理特征提取。在Brodatz纹理上验证后,将基于矩的纹理分析方法应用于CT肝脏扫描分类和前列腺超声分割。在我们的应用中使用了支持向量机和多通道主动轮廓模型。结果表明,我们的方法效果良好,但仍有一些局限性。