基于分层预测的高光谱图像分布式有损压缩算法  

Distributed Lossy Compression Algorithm for Hyperspectral Images Based on Hierarchical Prediction

作　　者：王林[1] 杨红[1] 卿粼波[1] 滕奇志[1] 何小海[1] 陈洪刚[1] WANG Lin;YANG Hong;QING Linbo;TENG Qizhi;HE Xiaohai;CHEN Honggang(School of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学电子信息学院,四川成都610065

出　　处：《无线电工程》2021年第11期1254-1260,共7页Radio Engineering

基　　金：国家自然科学基金资助项目(61871279,62081330105);中央高校基本科研业务费专项资金(2021SCU12061)。

摘　　要：针对高光谱图像星载压缩存在内存资源、计算能力有限等问题,提出一种基于分层预测的高光谱图像分布式有损压缩算法。该算法将高光谱图像分为K波段和WZ波段,并采用不同的编解码方式,其中K波段采用JPEG2000进行压缩,WZ波段采用分布式编码方案。利用小波变换去除高光谱图像的空间冗余,再利用高光谱图像的谱间相关性,设计出分层预测结构生成高质量的WZ波段边信息,以减少信道编码所需的码率。验证结果表明,与未考虑去除光谱冗余的静止图像编码标准JPEG2000、考虑了去除光谱冗余的传统高光谱图像有损压缩算法DWT+JPEG2000以及基于像素域的分布式有损压缩算法相比,该算法具有良好的率失真性能,在低码率情况下优势明显,适合星载高光谱图像有损压缩。In order to solve the problems of space-borne hyperspectral image compression,such as limited memory resources and computing power,a distributed lossy compression algorithm for hyperspectral images based on hierarchical prediction is proposed.The algorithm divides the hyperspectral image into K band and WZ band,and adopts different encoding and decoding methods.The K band is compressed by JPEG2000,and the WZ band uses a distributed coding scheme.Firstly,the spatial redundancy of hyperspectral images is removed by wavelet transform,and then the spectral correlation of hyperspectral images is used.Finally,a hierarchical prediction structure is designed to generate high quality WZ band side information,so as to reduce the bit rate required for channel coding.The experimental results show that compared with the still image coding standard JPEG2000 that does not consider the removal of spectral redundancy,and the traditional lossy compression algorithm for hyperspectral images DWT+JPEG2000 that considers the removal of spectral redundancy,and the distributed lossy compression algorithm based on the pixel domain,the proposed algorithm has good rate-distortion performance and obvious advantages under low bit rate conditions.This algorithm is suitable for lossy compression of spaceborne hyperspectral images.

关 键 词：高光谱图像 分布式信源编码 有损压缩 分层预测 边信息 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]