Remote Sensing Image Fusion Method Research Using Multi-wavelet Transform

As many kinds of satellite sensors sent successfully, Remote Sensing Image data collected from different sensors show variations in terms of temporal, resolution and frequency, with each bearing its own merits and short comings. In order to make the most use of these data, enable their inter complementing and image the target characteristics in a full scale, extracting useful information from each obtained image through certain arithmetic to fuse a new image has become a hot issue of the current study in this field. Such remote sensing image fusion aims to synthesize merit data gained from different sensors, as well as decreasing as much as possible or eliminate the uncertainty information.This paper selects proper Landsat-7 ETM+ remote sensing image data and then generates fusion images. To avoid the possibility that original images may affect the quality of the final fused image, this paper gives the remote sensing image data pretreatment as follow: using Lee filtering method to remove noise, using polynomial method to correct geometric emendation and introducing the image registration method based on the image Gray value; concerning several classical methods of pixel level such as linear average weighting method, high-pass filtering fusion method, IHS transform, principle component analysis transform, Pyramid fusion method, wavelet transformation method etc., contrasting and analysis is adopted to conclude the strengths and weaknesses of each and to identify their scopes of application.Based on the research of the characteristics of multi-wavelet transform and the typical methods of remote sensing images, Multi-wavelet is successfully applied into the process of Remote Sensing Image Fusion. And then a fusion method by means of CL balanced discrete multi-wavelet and IHS transform comes into being. And at the same time LBP operator based on local integration rule is used to analyze the components of high frequency, while average weighting rule is used to analyze the components of low frequency. This new method of fusing the multi-spectral images and panchromatic images can get high quality fusion images, which contains more detailed information, higher-definition and more obvious target characteristics information.