視頻分割和運動目標提取方法研究.pdf

1、安徽大學碩士學位論文視頻分割和運動目標提取方法研究姓名：涂錚錚申請學位級別：碩士專業(yè)：信號與信息處理指導教師：羅斌20070401安徽人學碩Jj學位論文視頻運動目標提取方法研究AbstractOwingtothefastdevelopmentofthecomputertechniques，itbecomesnotonlypossiblebutalsonecessarytoaccess，manipulate，andretrievethev

2、astaudio‘visualinformationbasedontheircontentsAtthesametime，newstandardsforcontent—basedcoding(MPEG4)anddescription(MPEG7)ofaudiovisualinformationarebeingdevelopedorfulfilledThus，itbecomesanurgentproblemtoextractanddescr

3、ibesemanticallymeaningfulmovingobjectsinavideosequenceAsanimpoaantresearchareainvideoandimageprocessing，theextractionofmovingobjectsinavideosequenceshasgreatvalueofresearchandapplicationThealgorithmsforextractingmovingob

4、jectsinvideosequencesarestudiedinthispaperFirst，themovingobjectextractionsystembasedonmotionareadetectionandspatio—temporalunionareproposed“l(fā)qaemainresearchworkwhichhavebeendoneinthisthesisisdescribedasthefollowing：(1)In

5、theaspectoftemporalmotionareadetection，animprovedhighorderstaffsticsalgorithmisproposedtOdetectmotionareaTheexperimentalresultsindicatethatthisalgorittmahasthesuperiorityofdenoisingcomparedwiththetraditionalhigh—ordersta

6、tisticsalgorithmsItCanmoreaccuratelyextractthemotionareafromcomplexbackground(2)ThroughapplyingthephasecongruencyedgedetectionalgorithmbasedonLogGaborcomplexwavelettransformtotheedgedetectionofframeimage，onepixelwidthedg

7、esofeveryframearedetectedFordifferentkindofvideoimages，thisalgorithmiscomparedwithsomeclassicalalgorithmsindepth，suchasCannyedgedetectionandwatershedsegmentationetcTheexperimentalresultsshowthattheproposedalgorithmisinse

8、nsitivetolowcontrastandilluminationvariationInaddition，theedgeofobjectsdetectedisone—pixelwidthandislocatedaccurately(3)Combiningimprovedhigh—orderstatisticsalgorithmwithaphasecongruencyedgedetectionalgorithmbasedOnLogGa