基于多波段预测的高光谱图像快速无损压缩算法  

Fast Lossless Compression Algorithm for Hyperspectral Images Based on Multi-band Prediction

作　　者：杨丽平[1] 胡红莉 朱福全[2] YANG Li-ping;HU Hong-li;ZHU Fu-quan(General Education Department,Sichuan Police College,Luzhou 646000,China;Dean's Office,Sichuan Police College,Luzhou 646000,China)

机构地区：[1]四川警察学院基础教学部,泸州646000 [2]四川警察学院教务处,泸州646000

出　　处：《价值工程》2022年第32期142-144,共3页Value Engineering

基　　金：四川省公安厅应用创新项目(编号:2021YYCX04);四川省泸州市科技局重点项目(编号:2019-GYF-14)资助。

摘　　要：为了实现对高光谱图像快速有效的无损压缩,本文利用维纳滤波理论建立了谱间多波段线性预测模型,提出了一种基于多波段线性预测的快速无损压缩算法。首先,利用中值预测器对第一波段进行谱内预测;然后,利用基于维纳滤波的多波段线性预测模型对其他波段进行谱间预测;最后,将预测产生的预测系数和残差图像进行算术编码。在AVIRIS 2006数据集上的仿真实验表明,该算法在16位校正图像、16位未校正图像和12位未校正图像上测试的压缩比分别达到4.71、2.77和4.90,平均压缩和解压缩时间分别约为151s和145s。该方法在保持较低计算复杂度的情况下有效提高了预测精度,在压缩效果和计算复杂度方面均优于其他算法。For fast and efficient lossless compression of hyperspectral images,a multi-spectral linear prediction model is developed using Wiener filter theory,a fast lossless compression algorithm using multi-band linear prediction is proposed in this paper.First,the median predictor is used to predict in first band,and then the multi-band linear prediction model based on Wiener filter is used to predict in other bands.Finally,the prediction coefficient and residual image are compressed by arithmetic coder.Experiments on AVIRIS 2006 hyperspectral images show that the algorithm yields an average compression ratio of 4.71,2.77 and 4.90 on 16-bit calibrated images,16-bit uncalibrated images and 12-bit uncalibrated images,the average times of compression and decompression are about 151s and 145s.Experimental results which effectively improves the prediction accuracy and outperforms other algorithms in terms of compression effect and computational complexity.

关 键 词：高光谱图像 无损压缩 线性预测 维纳滤波 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]