The Research On Fusion Methods Of Landsat-7 Remote Sensing Images Based On 1:50000 Relief Map Updating

As many kinds of satellite sending successfully. Remote Sensing data of multi-sensor, multi-temporal, multi-resolution, multi-frequency appears constantly and shows each merits and shortcomings. In order to use and develop those data completely, it is fused by some arithmetic to make up the insufficient information of single images. This paper selects proper data and then accepts fusion images, which improve line surface features information by effective fusion methods.By the analysis of image preprocessing having effect on image fusion, it can concludes that it is much better to resample multi-spectral images to the size of the panchromatic spectral image by doube-three cubed resampling, that it is better to have histogram matching for the panchromatic spectral image than not, that to have auto color gradation process for the fused images can add to the contrast in fusion images.There are corresponding evaluate criterions in different applications. The aim to this paper is to fuse images not only for improving the space resolution of the multi-spectral image but also for preserving the tone. In other words, on the basis of known low space resolution multi-spectral image and high space resolution panchromatic spectral image, it will simulate to create high space resolution multi-spectral image. To reach the aim, this paper designs a group of proper evaluation criterions on the basis of the qualitative and quantitative criterions, which is on the basis of visual interpretation, adopting average gradient and wavelet energy for analyzing the space resolution difference between fusion image and original undegraded multi-spectral image, adopting standard variance, standard deviation and covariance matrix determinant value for analyzing the difference of information between fusion image and original undegraded multi-spectral image, adopting mean, median and normalization least variance for analyzing the maintenance degree of spectrum information.There are 10 better image fusion methods when degrading the fused image to the half of original space resolution and then evaluating fusion images in qualitative and quantitative evaluation way. In order to give prominence to line surface features, this paper advances methods based on band selecting and primary component selecting, then fuses the selected band combination with panchromatic spectrum image by the 10 fusion methods, and finally educes that the best method is Daubiches 4 wavelet transform. From another point of view, the selected band images can improve line surface features and the method based on primary component selecting is much able to improve the contrast in water and surroundings.By above mentioned theories and experimental analyses, it concludes that the optimal image fusion steps are firstly, resampling multi-spectral images by double-three cubed method, secondly, selecting band 7,2,1 or primary component 1.3.4 to be the fused band combination, thirdly, having auto color gradation process for the fused images and having histogram matching for the panchromatic spectral image according to band combination, finally, fusing band combination with panchromatic spectral image by Daubiches 4 wavelet transform. There are experiments in other two scences ETM+ images of distinct areas for more tests on the effect of fusion steps, and the conclusion accords with above results to reach the applied aim. In the end. this paper simply describes the application of wavelet transform fusion methods based on band selecting and primary component selecting for relief map updating.