Research On Fusion Algorithm Of Multi-spectral And Panchromatic Remote Sensing Images

How to extract and integrate effective information reasonably from themultispectral and panchromatic image,in order to overcome the limitations of a singleimage itself, and finally achieve the goal of the accurate expression of the observationtarget, becomes the focus of the study. Wavelet transform,NSCT based multi-sourceremote sensing image fusion method is propsed in this paper.Firstly, based on the research background, significance and the present situation,remote sensing image fusion technology is discussed. The existing problem in fusionalgorithm and the future direction of the research are also pointed out.Furthermore,evaluation criteria of image fusion and several kinds of traditional fusion methods arepresented. Through the simulation experiment, analysis about the advantages anddisadvantages of the common fusion algorithm from the subjective and objective.Secondly, in-depth study on the wavelet analysis theory, based on wavelettransform, optimize the combination of commonly used fusion rules, the advantagesand disadvantages of each fusion rule obtained through the analysis of simulationresults. Then learned the model of Discrete Fractal Brown Random filed and learnedthe algorithm to calculate the fractal. Designed the fusion rules of high frequencycoefficient based the fractal dimension. Have achieved good results in enhancedspatial resolution and maintain the spectral characteristics of the fusion algorithm areverified by experiment.Finally, the àtrous wavelet transform and NSCT combined to form an imagedecomposition method based on àtrous-NSCT, and applied to the Multispectral andPanchromatic image fusion. The fusion rules in the high frequency coefficients basedon the fractal dimension has been improved.In the low-frequency coefficients,weighted fusion rules based on the sum-modified-Laplacian is established. think of thecharacteristics of Multispectral image and Panchromatic image, information entropy,clarity, similarity coefficient as the objective function to establish a multi-objective optimization model, use multi-objective optimization algorithm NSGA-II to solve themodel and the result as the weighting factor of low-frequency coefficients.The resultshows that the proposed method has a good performance on getting the spectrum andtexture information.