图像纹理分析的灰度-基元共生矩阵法

Image Texture Analysis by Grey-Primitive Cooccurrence Matrix

将描述纹理特征的统计方法和结构方法结合起来,提出灰度-基元共生矩阵,既能描述图像的灰度分布信息,又能表达局部细节。采用正规化处理,减少共生矩阵的计算量;将传统共生矩阵的一些特征如熵等引入基元—共生矩阵特征的计算,定义相应的特征向量。采用这种方法解决了单独使用某一方法的缺点,如计算量大、表达复杂、特征表达不够准确等。

This paper combines statistical methods with structural ones naturally. It introduces a gray-primitive co-matrix, which can be used to describe the gray distribution and the local detail structure in a textured image. In order to reduce calculation of co-matrix, a normalization to the matrix is used. Here some classical features, such as entropy, to co-matrix are introduced into the calculation to gray-primitive co-matrix. So it can use traditional ideas to solve new problem. Some feature vectors are defined to describe a textured image, which can be used to retrieve images from an image database. This method can avoid some drawbacks such as lots of calculation, inaccurate definition to the features, if one kind of methods is used in the image database retrieval.