简介:Anewhyperspectralimagecompressionmethodofspectralfeatureclassificationvectorquantization(SFCVQ)andembeddedzero-treeofwavelet(EZW)basedonKarhunen-Loevetransformation(KLT)andintegerwavelettransformationisrepresented.Incomparisonwiththeothermethods,thismethodnotonlykeepsthecharacteristicsofhighcompressionratioandeasyreal-timetransmission,butalsohastheadvantageofhighcomputationspeed.AfterliftingbasedintegerwaveletandSFCVQcodingareintroduced,asystemofnearlylosslesscompressionofhyperspectralimagesisdesigned.KLTisusedtoremovethecorrelationofspectralredundancyasone-dimensional(1D)lineartransform,andSFCVQcodingisappliedtoenhancecompressionratio.Thetwo-dimensional(2D)integerwavelettransformationisadoptedforthedecorrelationof2Dspatialredundancy.EZWcodingmethodisappliedtocompressdatainwaveletdomain.ExperimentalresultsshowthatincomparisonwiththemethodofwaveletSFCVQ(WSFCVQ),themethodofimprovedBiBlockzerotreecoding(IBBZTC)andthemethodoffeaturespectralvectorquantization(FSVQ),thepeaksignal-to-noiseratio(PSNR)ofthismethodcanenhanceover9dB,andthetotalcompressionperformanceisimprovedgreatly.