基于非监督学习的拼接图像盲取证算法分析-analysis of blind forensics algorithm for mosaic images based on unsupervised learning.docx

下载文档 降价啦

2
0
约4.05万字
约 59页
2018-06-05 发布于上海
举报
版权申诉
保障服务

基于非监督学习的拼接图像盲取证算法分析-analysis of blind forensics algorithm for mosaic images based on unsupervised learning.docx

1、本文档共59页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

基于非监督学习的拼接图像盲取证算法分析-analysis of blind forensics algorithm for mosaic images based on unsupervised learning

ABSTRACTDigitalimageforensicscanbeclassifiedintotwodifferentcategories,thedetectionofcopy-moveforgeryandimagesplicing,accordingtothedifferentoftechnologyusingintheforgeryprocess.Wepayourattentiontothedetectionofsplicinginthispaper.Sincewecan’tfindmoreaccurateinformationreflectingthefeaturesofsplicing,theexistingalgorithmsatpresenthavekindsoffaults.While，manyalgorithmsadoptsupervisedlearningmethodsinimageforensics,whichneedalargenumberoftrainingsamples,asaresult,thestabilityandefficiencyofthealgorithmsreliesheavilyonthestabilityofsomeimagecharacteristics.Since,theestablishmentofstandardsplicingimagedatabaseisnecessary.However,thiskindofdatabaseisveryfew,what’smore,itisnotperfect,andithasbeenimproved,thislimitstheapplicationofsupervisedlearningmethodsinimageforensics.Fromtheviewoftheformationprocessofdigitalimage,weproposeamorepracticalframeworkofsplicingforensicsystembasedonunsupervisedlearningmethods,whoseefficiencyisn’trelatedtothetrainingsamples.Inthispaper,thetemperedimageisregardedasacollectionofdifferentdatastructures,sotheforensicswillbeseenasaclassificationproblem.Inthatcase,imageauthenticidentificationdependsonlyontheimageitself,withouttheneedoftrainingsamples.Moreover,wegivetwodifferentkindsofalgorithmstooursystem,inordertopromotetheuseofimageforensics.Thethesiscompletesthefollowingworks:weanalysistheblockeffectinJPEGimage,aswhile,wegivetheselectingprincipleofblockingartifactmeasurement,what’smore,anenhancedblockingartifactmeasurementbasedonthisprincipleispresented,whichisusedinimagesplicingdetection,theoutcomeofexperimentshowsouralgorithmcandescribeamoreaccuratetemperarea.Weexplorethekurtosisconcentrationphenomenoninauthenticimage,whichisquitedifferentfromtemperone.Thus,weproposeacriterionthatdoesn’tdependonstandarddatabase,asaresult,thealgorithmismorestable.Welocatethetamperedpositionbyclassifyingthelocalnoiseeigenvalues.Kmeanalgorithmisagoodchoicebecauseofit’ssimpleandefficiencyinallclusteringalgorithm,although,Kmeansalgorithmissensitivetotheinitialclust