Research On Remote Sensing Image Fusion Algorithm Based On Multiresolution Analysis

This dissertation researches on the images of remote sensing fusion based on the achievement in related fields internal and abroad and image fusion algorithm based on multi-resolution analysis. The aim of it is to make the fused image be more consistent with human visual properties, highlighting the targets information, and to resolve the spectral distortion in image fusion problems of different types imaging systems, so as to provide a theoretical basis and practical methods for obtaining the information content of larger, wider application of image fusion.The key of imge fusion algorithm based on multi-resolution analysis is multi-resolution transform tools ans coefficients fusion rules. Image fusion algorithm of this dissertation uses the nonsubsampled Contourlet transform as the multi-resolution transform tools. Towards to different application areas, combined with human visual characteristics, regional characteristics of remote sensing images and the physical properties of imaging systems and other factors to develop coefficients fusion rules. Meanwhile, remote sensing images fusion algorithm based on human visual characteristics, region segmentation and characteristics of imaging systems are proposed separately.This dissertation researches on remote sensing images fusion based on human visual characristics for improving the human visual properties of fused images, which includes the infrared images fuse with visible images and multi-spectral images fuse with panchromatic images. The coefficient fusion fules of this two types image fusion algorithm are both developed with human visual function based on multi-resolution analysis, and this function is proposed on the foundation of traditional function of the human visual contrast.This dissertation researches on remote sensing images fusion algorithm based on region segmentation for giving prominence to target characristices of fused images according to region properties of remote sensing images. In the fusion algorithm of infrared and visible images, an infrared images segmentation algorithm is proposed firstly based on the local energy feature of it, and then a image fusion algorithm based on region segmentation is proposed by defineding the similarity of regional energy and regional structural. In the fusion algorithm of multi-spectral and panchromatic images, image segmentation algorithm of multi-spectral images is proposed firstly by combine with the Euclidean distance of spectrum, and then image fusion algorithm based on region segmentation is proposed by defineding regional mutual information estimate.In the applications of characteristics of imaging systems for remote sensing image fusion is known, the injection model of panchromatic images is established firstly according with multi-spectral imaging system model and general imag fusion (GIF) theory, meanwhile which combine with the spectral response function of each band of multi-spectral sensor for ALI plotting camera on EO-1 satellite, reflection rate of objects on earth and imaging sensor gain and bias in each band. And then a image fusion algorithm based on imaging system characristices is proposed by applying the model.Experimental results prove that remote sensing image fusion algorithm based on human visual characteristics can fully extract the edge information of the source images, so that the fused images have better contrast properties. Infrared and visible images fusion algorithm based on region segmentation not only retain the target characristics of infrared images, but also improve the edge features of fused images. Multi-spectral and panchromatic images fusion algorithm based on region segmentation fully maintain the spectral characristics of multi-spectral images and inject more detail information of panchromatic images into fused images. The fused images of multi-spectral and panchromatic images fusion algorithm based on imaging system characteristics have the minimal spectral distorition compared to other fusion algorithm, and more detail information of panchromatic images is also injected into fused images.