摘要
Stereomatchingisafundamentalandcrucialproblemincomputervision.Inthelastdecades,manyresearchershavebeenworkingonitandmadegreatprogress.Generallystereoalgorithmscanbeclassifiedintolocalmethodsandglobalmethods.Inthispaper,thechallengesofstereomatchingarefirstintroduced,andthenwefocusonlocalapproacheswhichhavesimplerstructuresandhigherefficiencythanglobalones.Localalgorithmsgenerallyperformfoursteps:costcomputation,costaggregation,disparitycomputationanddisparityrefinement.Everystepisdeeplyinvestigated,andmostworkfocusesoncostaggregation.Westudiedmostoftheclassicallocalmethodsanddividethemintoseveralclasses.Theclassificationwellillustratesthedevelopmenthistoryoflocalstereocorrespondenceandshowstheessenceoflocalmatchingalongwithitsimportantanddifficultpoints.Attheendwegivethefuturedevelopmenttrendoflocalmethods.
出版日期
2012年02月12日（中国期刊网平台首次上网日期，不代表论文的发表时间）
