IMD法在多光谱图像分类中的应用

数学形态学着重于分析和处理图像的几何结构信息。根据这一原理，在图像分类预处理过程中，对各种地物区域的几何结构特点进行分析，构造相应的结构元素，然后对图像进行形态迭代分解法(IMD，Iterative Morphological Decompostion)变换。在变换结果中，地物区域的灰度信息进行了归一化处理，同时区域的几何结构特点和独立地物得到了有效保留。将多光谱图像的IMD变换结果进行分类实验的结果表明，该方法可以有效提高多光谱图像分类的精度和效率，具有较强的适用性。

The Classification of Multispectral Images Using the IMD Algorithm

Mathematical Morphology emphasizes on analyzing and processing geometric structural information of images. Based on this theory, in the preprocess of image classification specific structuring element can be constructed with analyzing the features of geometric structure of different objects in images, and then images are to be transformed with the Iterative Morphological Decomposition (IMD) algorithm. So in the transformed images regional gray values are unified and geometric structural information of objects in images is effectively reserved. Results of multispectral images with IMD 's transformation are to be applied to a classification test, and the classification result shows that the IMD algorithm can effectively improve accuracy and efficiency and be of good applicability in application to classification of multispectral images.