基于稀疏分解的超光譜圖像壓縮算法研究.pdf

1、西安科技大學(xué)碩士學(xué)位論文基于稀疏分解的超光譜圖像壓縮算法研究姓名：徐喆申請學(xué)位級別：碩士專業(yè)：信號與信息處理指導(dǎo)教師：吳冬梅@Subject:HyperspectralImageCompressionBasedonSparseDecompositionSpecialty:SignalInfmationProcessingName:XuZhe(Signature)Instruct:WuDongmei(Signature)ABSTRACTH

2、yperspectralimageshavegreatapplicationsignificanceftheycanbeusedfthepixellevelevensubpixelidentificationeffectively.Howeveritshighresolutionratiotakeslargeramountofdataasacostwhichbringsabigproblemsftransmissionstage.Sot

3、heresearchofhighperfmancecompressionalgithmshasimptantapplicationsignificance.Thispaperproposedalossycompressionschemewhichusedsparsifysparsedecompositiontogether.What’smeitalsobuildedanothersystemascomparedwhichusedSPIH

4、TalgithmbasedonDWTaftersparsify.Firstlytheimage’ssparsitymeansithasfarmecontourinfmationthandetailed.Themodelusedinthispapercaneffectivelysparsethehyperspectralimages.Thetestdatashowsthatitmakessparserate(zeropixelratio)

5、increasedfromlessthan0.6%to55%above.Inadditionthesparseprocesscanalsoremovethespectralcrelationmethan94%.Secondlythepaperusedtwowaystocompressthesparedimages.Thefirstoneusedsignalsparsedecompositionalgithmtodecomposethem

6、thenusedfixedlengthquantizationmethodencodingtheatomsedbyMPalgithm.AnotheroneusedSPIHTalgithmbasedonlifted53wavelet.TheresultsshowthattheperfmanceofsparsedecompositioniswsethanSPIHTinnearlosslesscompression(PSNR40dB).But

7、inthehighratecompression，theadvantageofsparsedecompositionisobvious.TheaveragePSNRis31.12dBatthecompressionratioof102.40.Thisperfmanceisbetterthanmanyreferences’.Finallythepaperanalyzedthedifferencesofthetwoalgithmsdemon