高光谱遥感图像分辨率高，具有图谱合一的特性，现已得到广泛应用。但是高光谱图像中的混合像元问题广泛存在，严重制约着高光谱图像的分割效率，一直是遥感应用研究的难点和热点。目前常见的端元提取算法有：纯像元指数、交替最大体积法、最小封闭体积单形体等，这些算法可以从图像所有像元中提取纯光谱，但是提取速度慢、精度低。本文主要研究内容如下： 首先，本文引入简单线性迭代聚类的超像元分割算法（simple linear iterative clustering，SLIC），它是根据颜色和距离的相似度进行超像元分割，然后进行提取端元，大大提高了混合像元分解的速度，但其只能分割三个波段的图像，所以需要对高光谱图像进行降维压缩，从而降低了高光谱图像的精度。 然后，针对SLIC算法存在的问题，本文提出了高维简单线性迭代聚类的超像元分割算法（high dimension simple linear iterative clustering，H-SLIC），该算法对SLIC算法进行了改进，它是基于波段信息和位置信息的相似度进行超像元分割，并且该算法可以处理多波段的高光谱图像，无需对高光谱图像进行降维处理，大大提高了图像分割的精度。 最后，本文通过模拟数据和实际数据实验验证了SLIC和H-SLIC算法对混合像元分解速度和精度的影响，比较了不同参数下这两种算法进行混合像元分解的合理性和有效性，并通过光谱角和均方根误差对比验证了这两种算法，得出H-SLIC超像元分割算法克服了多波段高光谱图像分割的问题，提高了混合像元分解的效率，具备很好的实用价值。

Hyperspectral remote sensing image with high resolution has the characteristics of of Combination of imageray and spectrum. Now it has been widely used. However, the mixed pixel problem exists widely in hyperspectral images, which seriously restricts the efficiency of hyperspectral image segmentation. It is always a difficult and hot problem in the field of remote sensing applications. At present, the common algorithms for the extraction of the end elements are pure pixel index, alternating maximum volume, minimum volume enclosing simplex, etc. These algorithms can extract pure spectra from all pixels of the image, but the extraction rate is slow and the accuracy is low. The main research contents of this paper are as follows: First of all, a simple linear iterative clustering algorithm for super pixel segmentation Was introduced. It is based on the similarity of color and distance to the super pixel segmentation, then extracts the end. It greatly improved the speed of mixed pixel decomposition. But it can only deal with segmentation of three bands of image, therefore it is necessary to reduce the dimensionality of hyperspectral image, which reduces the accuracy of hyperspectral image. Then, according to the problems of SLIC algorithm, a super pixel segmentation algorithm based on high dimension simple linear iterative clustering algorithm was proposed in this paper, improving the SLIC algorithm. It is based on the similarity of band information and location information to super pixels segmentation. The algorithm can deal with multi band of the hyperspectral image, without reducing the dimension of the hyperspectral image, which greatly improved the precision of image segmentation. Finally, through the simulated data and real data experiments, the influence of the SLIC and H-SLIC algorithm on the speed and accuracy of mixed pixel decomposition was verified, and the rationality and validity of the two kinds of algorithm with different parameters for decomposition of mixed pixels was compared. Two kinds of algorithms were verified by SAD and RMSE. It is concluded that the H-SLIC algorithm overcomes the problem of multi band hyperspectral image segmentation and improves the efficiency of the decomposition of mixed pixels, which has good practical value.