Homologousfeaturepointextractionisakeyproblemintheopticalandsyntheticapertureradar(SAR)imageregistrationforislands.Anewfeaturepointextractionmethodusingathresholdshrinkoperatorandnon-subsampledwavelettransform(TSO-NSWT)foropticalandSARimageregistrationwasproposed.Moreover,thematchingforthisproposedfeaturewasdifferentfromthetraditionalfeaturematchingstrategiesandwasperformedusingasimilaritymeasurecomputedfromneighborhoodcirclesinlow-frequencybands.Then,anumberofreliablymatchedcoupleswithevendistributionswereobtained,whichassuredtheaccuracyoftheregistration.ApplicationoftheproposedalgorithmtoSPOT-5(multi-spectral)andYG-1(SAR)imagesshowedthatalargenumberofaccuratelymatchedcouplescouldbeidentified.Additionally,bothoftherootmeansquareerror(RMSE)patternsoftheregistrationparameterscomputedbasedontheTSO-NSWTalgorithmandtraditionalNSWTalgorithmwereanalyzedandcompared,whichfurtherdemonstratedtheeffectivenessoftheproposedalgorithm.Thealgorithmcansupplythecrucialstepforislandimageregistrationandislandrecognition.
