Multi-source Remote Sensing Image Registration And Fusion Research

Multi-sensors Remote Sensing Image Data Fusion (MRsIDF) is an important part of earth observing information acquisition and processing technique and is an efficient approach to cope with enormous and immense Remote Sensing (RS) image data. MRsIDF can be categorized to three levels: pixel (data), feature and decision. In order to improve MRsIDF at pixel level, the paper focuses on the following aspects:1. A systematic summary is given on the basic theory, characteristics, frame, methods, practice, history, actuality and problems of MRsIDF.2. The analysis and induction about the MRsIDF at pixel level are expatiated on the theory, data source, technique flow and algorithms.3. We introduced image fractal dimensions as an index to evaluate fusion image quality based on assessing the significance, performance, adaptability and other characteristics of the commonly used evaluation indexes. Experiment results show that, fractal dimensions is more stable, objective and accurate than other indexes in expressing image texture complication and image grayscale variation.4. A new method of relatively registering multi-sensors RS images at high accuracy has been proposed, which is based on several classic digital image matching algorithms and combines tiny facet digital differential rectification. A series of experiments show that, the method is sub-pixel accurate for registering multi-source, multi-resolution visible light RS images.5. In order to emphasize the linear objects in RS image, wavelet transform is inducted in image fusion and a multi-scale edges enhancement fusion approach is devised based on the multi -resolution analysis characteristics of the wavelet transform in this paper. The method fully utilized advantages of the multi-resolution analysis in detecting edges. Experiment results show that linear feature in the fusion image is well enhanced.6. Complex Wavelet Transform (CWT) can provide approximate shift invariance and better directional selectivity. We use CWT to fuse RS images and design another fusion algorithm based on CWT. Experiment results show that the fusion image based on CWT is generally better than the fusion image based on Real Wavelet Transform (RWT).7. The design and implementation for "Multi-sensors Remote Sensing Image Registration and Fusion System" software are finished in the project and can fulfill the primary application demands.