Study Of Multi-Sensor Image Fusion In Remote Sensing

In recent years, image fusion has been an important and useful technique for image analysis and computer vision. In the application of remote sensing, Image Fusion is such a procedure that image data sets of the same region which derive respectively from the sensors of different type are spatially registered, then the more predominant or the complementary information in these registered image data sets are organically combined into a new image data set via a certain algorithm, and the comprehension is performed on it. In comparison with the image data from a single information source, such new combined data can reduce or restrain the different meanings, indeterminacy, and error, utilizing the information coming from different sources furthest.The research object of this dissertation is focused on the fusion of Synthetic Aperture Radar and Multi-Spectral satellite images and with the purpose of optimizing the fusion of the integrated image information. The research includes Image Registration, Image Fusion and Evaluation Rules.Image Registration is the basis of Image Fusion. A novel feather based image registration algorithm is put forward in this dissertation. In the case that there exist great differences among the registered images, the higher performance is proved in the comparison experiments with the current image registration algorithms which base on feather in spatial domain.After systematically expounding the essential definitions, basic types, processing model, purpose and application of multi-sources RS image fusion, the thesis in depth researches the implementation of the wavelet multi-resolution analysis algorithm of several different types of images. The methods of image fusion based on the different wavelet analyzing technologies are brought out, and the performances of these methods are investigated, compared and analyzed, and some important conclusions come up. A novel image fusion algorithm based on un-decimated lifting scheme wavelet transform is presented in the dissertation. It has been proved by experiments that, in comparison with the existing remote sensing image fusion algorithm based on wavelet analysis, thisnovel algorithm can provide higher performance images on condition that there exist great characteristic differences among source images that are to be fused, especially for SAR and multi-spectral images. It can provide higher speed and alleviate the spectral degradation which usually goes with the existing algorithms, as well as constrain speckle noise of Radar images. In respect of image fusion algorithm, the fusion rules and operators are essential. A new set of fusion rules and corresponding operators are given in this thesis. These rules and operators associate the Mahalanobis Distance and rate of coefficient. They can be applied to different fusing methods.Another main contribution of the thesis is the objective evaluation of image fusion performance. Several evaluation criteria are presented in the dissertation for measure. These evaluation criteria are applied to compare the performance of various image fusion methods.A series of experiments on image fusion are given in the thesis and some worthy conclusions are obtained. The experiment and calculation results show that the fusion schemes and evaluation criteria are effectual.