基于FastICA的高光谱图像压缩技术  

Compression Technique for Hyperspectral Imagery Based on FastICA

作　　者：辛勤[1] 粘永健[1] 万建伟[1] 何密[1] 

机构地区：[1]国防科技大学电子科学与工程学院,长沙410073

出　　处：《电子科技大学学报》2010年第5期711-715,730,共6页Journal of University of Electronic Science and Technology of China

基　　金：国家自然科学基金(60572135);部级预研基金

摘　　要：提出了一种基于快速独立分量分析(FastICA)的高光谱图像压缩算法。首先引入虚拟维数算法估计图像中的目标端元个数,进而提取出感兴趣的目标端元矢量,并初始化快速独立分量分析的混合矩阵;利用最小噪声分量变换对原始数据进行降维,从降维后的主分量中提取独立分量;对独立分量进行恒虚警率检测与形态学滤波,实现目标分割。对高光谱图像进行谱间Karhunen-Loeve变换,利用比例位移法对感兴趣目标的小波系数进行提升,最后对各主分量进行最优码率的SPIHT压缩。实验结果表明,该方法在获得较高压缩性能的同时能够有效地保留感兴趣的目标。Efficient compression for hyperspectral imagery has been the research focus for the development of remote sensing technique.The small targets information protection during the compression process without any preknowledge should be necessarily considered.This paper presents a new lossy compression method for hyperspectral imagery based on fast independent component analysis（FastICA）.Virtual dimensionality is introduced to determine the number of target endmembers.The mixing matrix of FastICA is initialized by target endmembers.Minimum noise fraction is employed for dimensionality reduction of original data volumes,and FastICA is performed on the selected principal components to generate independent components.Then,constant false alarm rate detection is performed on each IC,which is followed by morphologic filtering.Karhunen-Loeve transform is used to decorrelate the spectral redundancy,general scaling-based method is selected to upshift the wavelet coefficients of interested targets.Finally,each principle component is allocated optimal rate and compressed by SPIHT algorithm.Experimental results on AVIRIS data show that the proposed method not only provides high compression performance,but also preserves targets interested effectively.

关 键 词：高光谱图像 独立分量分析 有损压缩 目标检测 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]