基于PCA变换的多光谱图像降维方法研究

Method of multi-spectral images reduce-dimensions based on PCA

图像处理技术已经在很多学科领域有着广泛的应用,其作用也日益凸显。相对于普通图像,高维图像数据量较大,信息繁多,对后续处理造成了诸多不便。因此,对高维图像处理的前期进行降维处理是很有必要的。文中对多光谱图像的降维方式进行了研究,采用了传统的主成分分析变换,利用矩阵的非线性变换,减少图像之间的相关性,对一部分噪声进行过滤处理,并将图像进行低通滤波,减小图像中的噪声,从而达到提高一定分类精度的实验结果。

Image processing has been widely used in different fields in our daily life, compared to other images, hyper-extraction images is bigger, with more data and contains more information, which made it difficult to analysis. So before the classification process, reduce image dimensions made it easier, faster, with less space to classify. This article has improved the mentioned a kind of image dimension-reducing methods, for the multi-spectral images, it uses the traditional PCA transformation first, the matrix transform to reduce the relations between each images, and also reduces a little noises. Since it considers little about noises, after the PCA transformation, uses low-pass filters to remove the noises in the images. This improved the classification accuracy.