基于双向预测的高光谱图像无损压缩  

Lossless Compression for Hyperspectral Images Based on Bi-Direction Prediction

作　　者：常进[1] 陶午沙 粘永健[3] 滕书华[3] 

机构地区：[1]南阳理工学院,河南南阳473004 [2]总装备部武器装备论证中心,北京100101 [3]国防科学技术大学电子科学与工程学院,长沙410073

出　　处：《电光与控制》2010年第10期65-67,89,共4页Electronics Optics & Control

基　　金：国家自然科学基金资助项目(60572135)

摘　　要：提出了一种基于双向预测的高光谱图像无损压缩算法。该算法首先采用自适应波段选择算法选出信息量较大的波段,然后利用聚类算法对这些波段的谱向矢量进行分类预处理。为了便于组织谱间预测过程,根据相邻波段相关性大小进行自适应波段分组,采用双向预测的方法去除谱间相关性。通过在参考波段和预测波段中定义三维上下文预测结构,在聚类结果的基础上,对各个像素分别训练最优的预测系数,从而实现当前波段的有效预测。对AVIRIS型高光谱图像的实验结果表明,该方法可获得较好的无损压缩性能。A new lossless compression algorithm for hyperspectral images based on bi-direction prediction was presented.The important bands which containing large amount of information can be determined by using the adaptive band selection algorithm,based on which clustering algorithm was used to sort the spectral vectors of these bands.To organize the inter-band prediction process efficiently,adaptive spectral band grouping was implemented to divide hyperspectral images into groups according to the correlation coefficients.Bi-direction prediction was introduced to eliminate the spectral correlation.3D contexts predicting structure were defined based on the current band and the reference band.On the basis of clustering results,the predicted coefficients of current pixel were trained in order to predict current band efficiently.Experiments on AVIRIS hyperspectral images showed that the proposed algorithm can obtain better lossless coding performance.

关 键 词：高光谱图像 无损压缩 谱向聚类 双向预测 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP751.1[自动化与计算机技术—检测技术与自动化装置]