Remote Sensed Imagery Fusion And Quality Evaluation

With the development of information technology, imagery fusion technology hasbeen widely applied in different fields. At the same time the massive remote sensedimagery also appeared. However, how to fuse so many kinds of remote sensedimagery is an imperative problem need to be solved, to which many scholars haveattached importance. Predecessors have made in-depth research to the technology ofremote sensed imagery fusion, which is divided into three levels: pixel level fusion,decision making level fusion and feature level fusion. But in the specific applications,it is difficult to design a method suitable for various fields. On the basis of theliteratures and summarizing previous research results, the algorithms of pixel levelfusion and its quality evaluation are researched roundly in the thesis, in which twosets of imagery are selected as experiment data, the main achievements are shown asfollows:(1).Expounding the basic conceptions of pixel level fusion detailedly, analyzingthe traditional methods of pixel level fusion and their characteristics comprehensively.(2).Selecting two sets of imagery of different resolutions for experimentalvalidation on the theory of pixel level fusion. For these two sets of imagery, HISfusion method is an important fusion method comparatively.On the basis of analyzinga series of imagery fusion methods, e.g. traditional HIS, wavelet, discrete wavelet,wavelet&HIS transform and HIS+DWT transform fusion are put forward.Experimental results show that wavelet+HIS transform is one of optimal methods, bywhich spectral information of TM imagery and spatial information of SPOT imageryare superimposed together to achieve the fused imagery of high quality. This methodcan not only reserve spectral characteristics of TM imagery perfectly, but also remainin spatial details of SPOT imagery to a large extent.(3).With the development of imagery fusion technology, contourlet transform, anew method of two dimensional imagery representation, is becoming our researchissue in the field of remote sensed imagery fusion. This method overcomes thelimitation of wavelet transform which cannot capture linear space entity description in2D space effectively, reserves the relationship of time and frequency of wavelet in onedimensional space, achieves wavelet factorization in arbitrary direction at anyscale.Also, contourlet transform is prone of describing the image contour and textureinformation to compensate for wavelet transform algorithm. The method is applied toremote sensed imagery fusion, spatial details in high frequency field are factorized toeach direction to promote the imagery fusion choices,obtain more fine fusion impact.On the basis of contourlet transform, imagery fusion algorithm of contourlet+HIStransform is put forward in this thesis. Experiment results show that compared withthe wavelet+HIS transform and IHS+DWT transform, this algorithm enhances the spatial details of imagery, improves spatial resolution of the multispectral imagery,and achieves the fused imagery of high quality both in the subjective and in objectiveaspects.(4).Evaluating the quality of fused imagery, analyzing the indices of qualityevaluation comprehensively. Research results prove that the mixed fusion method issuperior to the traditional fusion methods, also, it is a beneficial attempt in the field ofimagery fusion technology.