Research On Multi-Source Image Fusion Algorithms

Multi-source image fusion is an important and useful subject in Image Understandingand Computer Vision. In recent decades, image fusion method in wavelet domain isbecoming a most important field. In this dissertation, the fusion algorithms of waveletdomain are studied in details. The main contributions are as follows:1. A systematic overview of Multi-source image fusion theory and methods ispresented. The development of image fusion in wavelet domain is introduced andthe problems are pointed out.2. The image fusion method based on Pyramid is studied. A new dynamicmulti-scale gradient pyramid modulation fusion method is presented, which canbe used to analyze multispectral remote sensing image. Experimental resultsshow that comparing with traditional contrast based method and wavelet basedmethod, our method can reserve and strengthen significant features of variousspectral images.3. The image fusion methods based on M-band wavelet transform and Multiwavelettransform are studied. Considering both the multi-scale multi-directioninformation and the correlation property in M-band wavelet domain, an imagefusion scheme based on Cross-band-window-Intrascale(CBWI) is put forward.The simulation results in multi-focus image fusion show that our method issuperior to the method based on 2-band wavelet in both the subjective andobjective evaluation. A translation-invariant extension of multiwavelet transformcalled stationary multiwavelet transform is proposed, which yields anovercomplete signal representation, and consequently incorporates the advantagesof multiwavelet transform and translation-invariant property. A new image fusionalgorithm based on stationary multiwavelet transform is presented.4. The image fusion method based on 2-band lifting wavelet is studied. The recentdevelopment of the lifting wavelet theory is discussed. Taking advantage of itscharacteristics such as simple, fast and small computering memory, a new liftingwavelet based image fusion method for SAR and CCD is presented. Experimentsunder various scenes like shore, searshore and plain demonstrate the effective ofthe method.5. Multiple Sensors image fusion method is studied. A false color image fusionalgorithm based on HSI color space for IR and CCD sensors is proposed. Compared to the gray fused image, the false color fusion result not only canreserve space resolution rate, but also can stand out some special informationsuch as temperature.6. The prospect of image fusion in wavelet domain is analyzed, and some possibleapplications of wavelet domain model, M-band multiwavelet system, ridgelet andcurvelet theory in image fusion are given.