研究目的是对代数多重网格(AMG)方法进行分析,粗网格中会保留强连接部分而去掉弱连接部分,可以提取图像的纹理信息。将AMG方法提取的图像的纹理特征结合到图分割算法中,针对具有纹理特征的图片构建能量函数,并使用最大流方法进行优化。使用一些自然图像进行了验证,结果证明针对该方法能够较好地提取图像的纹理特征。
Algebraic multi-grid method is analyzed and is applied in the normalized cut method to extract me texture ieature of the image. Large grid density appears in the image regions with radical changes, and small one in the smoother regions. Singulari- ties in the image can be detected by the AMG method, and especially the singularities in the texture image. An energy function is constructed for the texture feature and is minimized using max-flow method. Experimental results show that the proposed method can extract more texture details.