基于块稀疏表达模式的高光谱图像压缩  

Hyperspectral image compression based on block sparse representation patterns

作　　者：种衍文[1] 郑炜玲 潘少明[1] Chong Yanwen;Zheng Weiling;Pan Shaoming(State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, Chin)

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《华中科技大学学报（自然科学版）》2017年第12期60-65,72,共7页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61572372;41671382;41271398)

摘　　要：针对高光谱图像压缩算法存在的解码端计算复杂度高,且没有充分考虑高光谱图像结构特征信息等问题,提出了一种基于块稀疏表达模式的高光谱图像压缩方法.主要通过在编码端利用结构字典对稀疏系数进行结构化压缩编码,避免解码端非线性重构,以达到缩短高光谱图像重构时间的目的.实验证明该方法在压缩比较低(0.015 9)时依然能获得较高的重构精度(峰值信噪比为22.240 3,结构相似度为0.511 4).To solve the problem existing hyperspectral image compression schemes have high computational complexity at decoder and does not take the structural feature information into full consideration,a new compression approach based on block sparse representation patterns was proposed.The sparse coefficients were encoded over structural dictionary at encoder to avoid non-linear reconstruction at decoder,in order to shorten the reconstruction time.The results show that in the case of low compression rate（0.015 9）,the proposed approach can still guarantee the reconstruction accuracy（peak signal to noise ratio is 22.240 3,structural similarity is 0.511 4）.

关 键 词：高光谱图像压缩 结构化压缩感知 结构字典 稀疏表示 块稀疏表达模式 

分 类 号：TN919.81[电子电信—通信与信息系统]