深度学习算法的多光谱图像压缩方法  

Multi-spectral image compression based on deep learning algorithm

作　　者：徐景秀[1] 张青[1] XU Jingxiu;ZHANG Qing(Huanggang Normal University,Huanggang Hubei 438000,China)

机构地区：[1]黄冈师范学院,湖北黄冈438000

出　　处：《激光杂志》2020年第12期73-77,共5页Laser Journal

基　　金：湖北省自然科学基金(No.2014CFC1100)。

摘　　要：多光谱图像压缩可以节约存储空间,加快数据传输速度,当前多光谱图像压缩方法存在损失有效数据严重,压缩后多光谱图像无法反映原始多光谱图像的信息,为了获得更加理想的多光谱图像压缩效果,提出基于深度学习算法的多光谱图像压缩方法。首先分析当前多光谱图像压缩算法的研究进展,找到易引起多光谱图像压缩效果的原因,然后采集大量多光谱图像,去除多光谱图像中的噪声,初步减少多光谱图像存储空间,最后引入深度学习算法对多光谱图像中的冗余信息进行检测,删除大量的冗余信息,实现多光谱图像压缩。测试结果表明,深度学习算法可以高效地对多光谱图像进行压缩,多光谱图像压缩比率要高于其它方法,减少了多光谱图像存储所需要的空间,提高了多光谱图像传输速度,缩短了多光谱图像压缩时间,验证了深度学习算法用于多光谱图像压缩的优越性。multi-spectral image compression can save storage space and speed up data transmission.At present,the multi-spectral image compression method has serious loss of effective data.After compression,the multi-spectral image cannot reflect the information of original multi-spectral image.In order to obtain more ideal compression effect of multi-spectral image,a multi-spectral image compression algorithm based on depth learning algorithm is proposed.Firstly,the research progress of multi-spectral image compression algorithm is analyzed to find out the reasons that are easy to cause the compression effect of multi-spectral image.Secondly,a large number of multi-spectral images are collected to remove the noise in the multi-spectral image and initially reduce the storage space.Finally,the deep learning algorithm is introduced to detect the redundant information in multi-spectral image and delete a large number of redundant information,so as to realize the multi-spectral image compression.The test results show that the deep learning algorithm can effectively compress multi-spectral image,and the compression ratio of multi-spectral image is higher than other methods,which reduces the space needed for multi-spectral image storage,and is conducive to improving the transmission speed of multi-spectral image,and the compression time of multi-spectral image is shorter,which verifies the superiority of the deep learning algorithm in multi-spectral image compression.

关 键 词：多光谱图像 存储容量 压缩比率 数据冗余 深度学习 

分 类 号：TN298[电子电信—物理电子学]