Research On Multisensor Image Fusion Algorithoms

Multi-sensor image fusion is to integrate complimentary and redundant information from multiple sensor images to get a more accurate and more reliable image than any of the individual sensor source images. Image fusion can complement each other's advantages from multiple images and reduced the amount of data by only reserve the fused images. Image fusion has broad prospects in the military and civilian applications, is a hotspot in the image processing field.Image fusion can be divided into pixel-level fusion, feature-level fusion and decision-level fusion according to the abstraction measure of image information. The pixel-pixel level fusion is the base of the feature-level fusion and the decision-level fusion, it can provide more extensive details than other fusions and is beneficial for further image analysis and processing. The research in this paper is focused on the pixel-level image fusion.Firstly, the basic concept and the level for image fusing are described, the research background and significance development and research focus are introduced, then summarize the classification and common evaluation indexes of pixel-level image fusion. Second, the basic principles and fusion algorithms based on wavelet transform are introduced, and a variety of fusion methods are deeply studied and simulated.Considering to retaining more source image information and ensure better edge feature, a fusion method based on texture energy -based is presented. Simulation results show that this algorithm can improve the quality of the fused image effectively. In addition, several common algorithms for the fusion of panchromatic images and multi-spectral images are studied. A modified method based on Brovey transform is proposed in view of the spectrographic inconsistency between panchromatic image and multi-spectral image. Simulations show that the proposed can make full use of the spectrographic information of the original multi-spectral images on the premise that the spatial resolution of fused image has been improved. Especially in the darker areas of original scene, the highlighted information greatly supported the identification and classification of ground objects. Finally, summarize the work of this thesis and point the research direction in future.This thesis selects some source images to carry out the experiments of image fusion, draw some significant conclusions. The main contributions of this paper are summarized as follows:(1)Two image texture templates for texture detection from multi-direction are defined. The simulation show that the texture templates can detecting multi-direction texture distribution effectively;(2)An image fusion algorithm based texture energy in the wavelet domain for multi-focus image fusion is proposed. This method can keep more information from source images, and improve the clarity of the fused image;(3)This thesis presents a method based spectral correction for the fusion of multi-spectral images. This method is a modified algorithm of Brovey transformation fusion. Simulation improve this method can make full use of the spectral information from the original multi-spectral image, highlight the surface features of the ground objects. The fusion image is more suitable for identification and classification of objects.