学术
一种稳健的高光谱图像压缩方法
苏令华[1,2] 李纲[1] 衣同胜[3] 万建伟[1]
[1]国防科技大学电子科学与工程学院,湖南长沙410073 [2]空军大连通信士官学校,大连116600 [3]92941部队,辽宁葫芦岛125000
摘 要:
提出了一种稳健的高光谱图像压缩技术。估计原始数据的虚拟维数,继而采用一种非监督的正交子空间投影技术提取端元矢量,构成矩阵作为快速独立分量分析的初始化混合矩阵,进行ICA变换。对变换后的独立分量图,采用二维分层树集合分裂进一步压缩编码。计算机仿真结果表明,该压缩算法在取得40∶1压缩比的同时,能有效地保持数据,尤其是小目标和异常的谱向特征,是一种高效的三维数据压缩方法,可有效地应用在小目标检测及异常检测中。[著者文摘]
文章出处:
《光学精密工程》-2007年15卷10期 -1609-1615页
栏目信息:
文献标识码:
A
文章编号:
1004-924X（2007）10-1609-07
A robust hyperspectral image compression method
SU Ling-hua , LI Gang , YI Tong-sheng , WAN Jian-wei （1. College of Electronic Science and Engineering , National University of Defense Technology, Changsha 410073,China; 2. Dalian Sergeant School of Communication o fAir Force ,Dalian, 116600;3. 92941 Unit of PLA, Huludao 125000,China）
Abstract:
A robust hyperspectral image compression method is proposed for the detection of small target and abnormity. The virtual dimensionality of original data is estimated, then the endmember vectors are extracted using an unsupervised orthogonal subspace projection method. Fast-ICA transformation is performed with the endmember vectors as initialization matrix, and the IC images are coded by 2D Set Partitioning in Hierarchical Trees （SPIHT）. The experiment results show that the algorithm has a high compression ratio （40 ： 1）, while the spectral characteristics are preserved. This algorithm is an efficient 3D compression method for the detection of small target and abnormity.[著者文摘]
Key words:
image compression; hyperspectral image; independent component analysis~ initialization
基金资助:
国防预研基金资助项目（No.41321090202）
更多>>免费文章
cqvip.com