Abstract: In the process of the texture image recognition and classification，many images which have the same texture features usually show the diversity of direction.These images are supposed to be the same category.So the method how to get the rotation invariant features is needed.There are many methods to solve it，such as geometric moments，orthogonal moments，gray level co-occurrence etc.But the first two methods are time consuming，and the performance of the third is not satisfying.A method based on Gray-Gradient Co-occurrence Matrix is recommended to get the rotation invariant features.Furthermore，a fast algorithm for computing the Gray-Gradient Co-occurrence Matrix is proposed.Experimental results demonstrate that the method is very efficient for getting the rotation invariant features，and the fast algorithm has largely reduced the computation.

Key words: texture analysis, gray level co-occurrence matrix, rotation invariance, gray-gradient co-occurrence matrix

摘要： 在对纹理图像进行分类识别过程中，许多具有相同纹理特性的不同图像经常在方向上呈现多样性。这些图像应该被归为一类。针对这一问题，有许多方法可以得到旋转不变性特征，例如：几何矩，正交矩，灰度共生矩阵等，然而，前两种方法计算量很大，第三种方法效果也不令人满意。提出了一种基于灰度-梯度共生矩阵的方法来得到旋转不变特征量，并且提出了一种快速计算灰度-梯度共生矩阵的算法。实验表明利用灰度-梯度共生矩阵的方法得到旋转不变量的方法非常有效，快速计算灰度-梯度共生矩阵的算法也大大减小了计算量。

关键词: 纹理分析, 灰度共生矩阵, 旋转不变, 灰度-梯度共生矩阵