Remote Sensing Image Fusion Method Research Based On Wavelet Transform

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 by double-three cubed resampling, that it is better to have histogram matching for the panchromatic spectral image than not.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 energy for analyzing the improvement degree of the space resolution, adopting standard variance and covariance matrix determinant value for comparing the maintenance degree of image information, adopting mean for measuring the maintenance degree of image spectrum resolution, adopting entropy for measuring the plentiful degree of image information, adopting correlation coefficient for evaluating the image fidelity.This paper shows some better image fusion methods, and then evaluates fusion images in qualitative and quantitative evaluation way, and finally educes that the best method is wavelet transform. In order to give prominence to line surface features, this paper proposes wavelet transform method based on band selecting, and then proposes a controllable method based on wavelet transform method. This method provides controllable parameter, in which images are fused in wavelet space. So the balance between spectral characteristics preservation and high spatial resolution retention can be achieved at different levels. By above mentioned experimental analyses and comparison, it concludes that the method that resampling by double-three cubed method, and then having histogram matching for the panchromatic spectral image, selecting band 7,4,1 to be the fused band combination by wavelet transform, is the best result at the angle of qualitative and quantitative analysis.A new controllable fusion method based on wavelet transform makes some good improvement for wavelet transform fusion method at some extent.