A New Image Mosaics Algorithm Based on Feature Points Matching gg g
Guandong Gao, Kebin Jia · 2007
Imagemosaicingtechniqueisusedincreatingvast� sceneimagesandpanoramicimages.� Ithaswide� applicationvalueintheculturalheritageprotecti andvirtualrealityfield.�Anautomaticalgorithmb ased� onfeaturepointsofimagesispresentedinthispa per.� Firstly,� featurepointsareextractedusingHarris� cornerdetectortoperformcorrelation.�Then,�anew � methodoffilteringformatchingpointsispresente dby� choosingpairsofcorrelationfeaturepointswith� clusteringalgorithmbeforepurifyingthemwith� RANSACalgorithm.�Afterprojectiontransitionmatri x� isworkedout,� imagesareprojectedtoacylinder� surface.�Finally,�inordertodealwithghosting,�a �fast� methodincludingoptimizationpathandluminance� weightingfunctionisusedtoobtainagoodresult. � Experimentalresultsshowthatthisisarobustima ge� mosaicsalgorithm,� andafterchoosingdatawith� clusteringalgorithmbeforehand,� theefficiencyis� greatlyenhancedandmakesthealgorithmmore� suitableinpractice.�