论文部分内容阅读
在图象处理与识别的广泛实际问题中,如遥感图象的分析中,不仅要求我们给出图象上所研究对象的形状,而且还要求我们给出对象区域的内部信息结构或性质,如此,则必将涉及到对象区域内部的纹理及其纹理特性,因此,如何自动的进行纹理的分析和分类乃是图象处理与识别问题中的一个十分重要的课题。本文将采用Haralick等人所提出的空间灰度共现矩阵这一有效的方法来分析和描述纹理;然而,它的问题是对同一图象反映其纹理的共现矩阵并不是唯一的,为此,我们必须给出一个共现矩阵结构性能的度量标准,以便获得反映图象纹理结构的最好的共现矩阵;最后,在如此选择的共现矩阵上,去选取适宜的纹理特性,并利用最大似然方法进行纹理分类。
In a wide range of practical problems of image processing and recognition, such as the analysis of remote sensing images, we are asked not only to give the shape of the object under study, but also to give us the internal information structure or nature of the object area, , It is bound to relate to the texture and texture features inside the object area. Therefore, how to automatically analyze and classify textures is a very important issue in image processing and recognition. However, its problem is that the co-occurrence matrix that reflects the texture of the same image is not unique, , We must give a measure of the performance of the co-occurrence matrix structure in order to get the best co-occurrence matrix that reflects the texture structure of the image. Finally, on the co-occurrence matrix thus selected, we select the appropriate texture characteristics and use Maximum likelihood method for texture classification.