论文部分内容阅读
大尺寸图像按照区域类别分块,用灰度共生矩阵法对各子块提取出纹理特征,并利用类似Hu的不变矩的方法对各子块纹理信息进行统计,从而提取出与大图像中各纹理的空间位置相关的一组纹理特征值。该方法对传统的共生矩阵法进行了改进,使其适用于大尺寸图像多区域类别的纹理特征提取,可用于大尺寸遥感图像的基于内容的检索和初分类研究。
The large-size images are divided into blocks according to the region type, and the texture features are extracted from each sub-block by the gray-level co-occurrence matrix method. The texture information of each sub-block is statistically calculated by a method similar to Hu’s invariant moment, A set of texture eigenvalues related to the spatial location of each texture. This method improves the traditional co-occurrence matrix method and makes it suitable for the texture feature extraction of large-scale multi-region images. It can be used for content-based retrieval and initial classification of large-scale remote sensing images.