The Research On Distributed Source Coding Of Hyperspectral Images

With the growing requirement of ground resolution and spectrum resolution for hyperspectral remote sensing image,the related amount of information required is also increasing sharply.However,as the the transmission channel of space communication,especially deep space communication has characteristics of heterogeneous,time-varying,limited bandwidth,high bit error rate and so on,experts are facing big challenges before compression and transmission of hyperspectral remote sensing image.Now it is affecting and limiting the practical application of the image.In order to resolve the problem,experts began to pay high attention to compressing technology of hyperspectual image,the technology itself now has been an important forefront issues in field of hyperspectral remote sensing image processing field.This paper presents a hyperspectral remote sensing image compressing technology based on distributed source coding,and the above problem can be effectively solved with it.First of all,the paper introduces the overview of remote sensing imaging technology,mainly analyzes on research status and development trend of multispectral and hyperspectral images,and respectively introduces three coding algorithms meanwhile,they are prediction based coding method,transformation based method and vector quantization based coding method.Secondly the paper introduces the background and current situation of distributed source coding method,Slepian-Wolftheory(for lossless compression),Wyner-Ziv theory(for lossy compression)and three traditional distributed coding methodsdetailedly.At last,this paper proposes two hyperspectral remote sensing image coding schemes based on above theories:(1)Lossless distributed source coding of hyperspectral images based on quadtree segmentation.This scheme has enhanced the adaptability of formed coset code and achieved better compromise between compression effect and coding efficiency through self-adaptive quadtree segmentation.(2)Lossy distributed source coding of hyperspectral images based on ROI(Region of Interest).This scheme firstly extracts ROI from the key frame,then perform JPEG-LS lossless coding on ROI,SPIHT coding on background region and compressive sensing based coding method on WZ frame.Besides their well reserving the advantages of distributed coding method,the two proposed methods in this paper also have different ideas.The first arithmetic commendably accomplishes demand of lossless coding in the environment of lower coding complexity through self-adaptive blocking the images.The second arithmetic achieves lower codingcomplexity,stronger ability of error control and quality scalability of bit stream through adaptively handling for K frame and WZ frame.