基于独立分量分析的高光谱图像降维与压缩算法

论文提出了一种基于快速独立分量分析的高光谱图像降维算法.利用虚拟维数算法估计需要保留的独立分量数目,采用非监督端元提取算法自动获取端元矢量,并对快速独立分量分析的混合矩阵进行有效初始化.采用最大噪声分离变换对原始数据进行预处理,利用快速独立分量分析从变换后的主分量中依次提取出各端元对应的独立分量,最后对各个独立分量分别实施无损压缩.实验结果表明,该算法降维后的独立分量具有较好的地物分类性能,并且可以获得较好的压缩性能.

Dimensionality Reduction and Compression of Hyperspectral Imagery Based on Independent Component Analysis

This paper presents a fast independent component analysis（fastICA） approach to dimensionality reduction for hyperspectral imagery.Virtual dimensionality is introduced to determine the number of independent component to be retained.The mixing matrix of fast-ICA is initialized by endmember vectors extracted from hyperspectral imagery by using unsupervised method,which can resolve the random order of independent components.Maximum noise fraction is used for preprocessing of original data,and fastICA transformation is performed on the selected principal components to generate independent components according to the acquired endmembers.Finally,the independent components are compressed losslessly by JPEG-LS.Experimental results show that the proposed algorithm can achieve better coding performance,the independent components after dimensionality reduction can also provide better object classification results.