基于独立成分分析的图像颜色特征提取研究

Color-feature extraction based on independent components analysis

结合独立成分分析(Independent Components Analysis,ICA)和最大相似度分类器(Maximum Likelihood)的特点,本文提出了一种基于ICA的多频谱遥感图像色彩分类的算法。该算法提取图像的色彩的独立成分,去除了图像的R、G、B之间的相关性,光谱独立成份用来聚集像素,使用Maximum Likelihood对像素进行颜色分类。实验结果表明,该方法识别性能好,准确度高,是对多频谱遥感图像的颜色特征提取的一种有效方法。

This article propose a algorithm of ICA to extraction color feature of remote sensing Image. The algorithm combines the advantage of ICA and Maximum Likelihood. The algorithm extracts the spectral independent components of muhispeetral remotely sensed images and move the correlation of R,G and B. Spectral independent components are used to cluster pixels using Maximum Likelihood to classify the pixels. Experimental results show that the algorithm improve the classification performance of muhispectral remotely sensed images.