Characterisation ofImageFusion Quality Metrics forSurveillance Applications overBandlimited Channels

Image fusion isfinding increasing application inareas suchasmedical imaging, remote sensing ormilitary surveillance us- ingsensor networks. Manyofthese applications demand highly com- pressed datacombined witherror resilient coding duetothecharac- teristics ofthecommunication channeL Inthis respect, JPEG2000 hasmanyadvantages overprevious image coding standards. This paperevaluates andcompares quality metrics forlossy com- pression using JPEG2000. Three representative image fusion algo- rithms: simple averaging, contrast pyramid anddual-tree complex wavelet transform basedfusion havebeenconsidered. Numerous in- frared andvisible test images havebeenused.Wecompare these results with apsychophysical study where participants wereasked to perform specific tasks andassess image fusion quality. Theresults showthat there isacorrelation between mostofthe metrics andthepsychophysical evaluation. Theyalso indicate that selection ofthecorrect fusion methodhasmoreimpact onperfor- mancethanthepresence ofcompression.