Fusion Algorithms Of Remote Sensing Images At Pixel-level

The thesis focuses on remote sensing image fusion algorithm at pixel-level, especially the multispectral and the high space-resolution panchromatic images fusion algorithm in transform region. With image multi-resolution analysis,we can get the high-frequency detail information from the low-frequency approximation information, then, the different information images are fused with different fusion rules, in order to get the target of fusing MS image and Pan image, which is preserving spectral qualities. The main work can be summarized as follows:1. Chapter 1 depicts the background, objective and meaning of remote sensing image fusion algorithm. The thesis is concerned with the multispectral and the high space-resolution panchromatic images fusion methods in transform region. So, the paper builds the general framework for remote sensing image fusion in transform region on the based of G Piella'module of the framework. Moreover, the paper improves a matching expressions.Also, some existing objective fusion performance measures are introduced because the evaluation of fusion result is an important component of image fusion.3. Based on the IHS transform, Mallat wavelet transform and the method in literature [14] ,a novel remote sensing image fusion approach with Mallat wavelet transform is presented. First, the low-frequency approximation images of the input images which obtained by using the Mallat wavelet transform are fused based the methods of averaging the approximation,and the high-frequency detail images are fused by the new rules,which uses the region-difference as activity measure, the improved correlation as match measure. Finally, the fused image is reconstructed using the inverse transform. The experimental results show that the new fusion method can achieve better fusion performance than energy fusion method in literature [14].4. because the Mallat Wavelet Transform has two main disadvantages: lack of shift invariance and poor direction selectivity for diagonal features and the Dual-Tree Complex Wavelet Transform owns many excellent properties: approximate shift invariance, good directional selectivity, perfect reconstruction, and limited redundancy, it is used in this thesis .The thesis introduces the construction and properties of the DT-CWT and brings up a new image fusion algorithm with DT-CWT. The algorithm employs the method of averaging the approximation to get the low-frequency approximation images of the input images which obtained by using the DT-CWT and adopts the rules which gets the grads as activity measure, the improved correlation as match measure to get the detail coefficients. Due to the algorithm employs grads as activity measurewhose activity is higher than region-difference and make the rate of account becoming higher, it comes to a better method . Furthermore, due to shift invariance property and good direction selectivity, the dual-tree complex wavelet transform can preserve edge information and texture information. So the method proposed in this paper can obtain good performance than Mallat wavelettransform in the same condition.5. On the based of introducing Contourlet transform, the thesis imports Nonsubsampled Contourlet Transform (NSCT). The NSCT is a complete shift-invariant directional multiresolution image representation, and it can retain the same dimension of the images of different levels which are obtained by the NSCT, so it can get more information in different levels by appending method. Based on the appending thought and fusion method in literature [37], the thesis propose a new method. First, the algorithm uses NSCT to represent the input images at different scales. Then it employs different fusion rules to deal with different images information. The approximation images are processed by averaging the approximation of the input image and the high-frequency detail images in different levels are processed by different fusion rules. The low yield high-frequency detail information is fused by using the grads as activity measure, the improved correlation as match measure and the high yield high-frequency detail information is fused by obtaining the maximum of the MS image and Pan image decompounded by NSCT on the same level.Then, we obtain the composite image by using the inverse transform. Finally, the paper compares the result of method presented in this thesis with others. Although the algorithm increases the computation, the experimental results indicate that images have good qualities.6. In the end, the paper summarizes the research work, and point out the NSCT is a good image multi-resolution analysis tool; different images information which obtained by using the multi-resolution analysis tools should be dealed with by different fusion rules;the improved image fusion algorithm with NSCT mentioned in the thesis can make the fusion images preserve the spectral qualities as much as possible while improving the space resolution quality.The problems and disadvantage of current research work are also analyzed in this thesis. Moreover, the paper analyzes the problems and disadvantage and discusses the recommendations for further work for remote sensing image fusion.