简介:Stripenonuniformityisverytypicalinlineinfraredfocalplane(IRFPA)anduncooledstarringIRFPA.Wedeveloptheminimummeansquareerror(MMSE)methodforstripenonuniformitycorrection(NUC).ThegoaloftheMMSEmethodistodeterminetheoptimalNUCparametersformakingthecorrectedimagetheclosesttotheidealimage.Moreover,thismethodcanbeachievedinoneframe,makingitmorecompetitivethanotherscene-basedNUCalgorithms.Wealsodemonstratethecalibrationresultsofouralgorithmusingrealandvirtualinfraredimagesequences.Theexperimentsverifythepositiveeffectofouralgorithm.
简介:Inlandmark-basedimageregistration,estimatingthelandmarkcorrespondenceplaysanimportantrole.Inthisletter,anovellandmarkcorrespondenceestimationtechniqueusingmeanshiftalgorithmisproposed.Imagecornerpointsaredetectedaslandmarksandmeanshiftiterationsareadoptedtofindthemostprobablecorrespondingpointpositionsintwoimages.Mutualinformationbetweenintensityoftwolocalregionsiscomputedtoeliminatemis-matchingpoints.Multi-levelestimation(MLE)techniqueisproposedtoimprovethestabilityofcorrespondingestimation.Experimentsshowthattheprecisioninlocationofcorrespondencelandmarksisexact.Theproposedtechniqueisshowntobefeasibleandrapidintheexperimentsofvariousmono-modalmedicalimages.