Based On The Multi-source Image Fusion Method For Sampling Contourlet Transform Under Study

The purpose of image fusion is to take full advantage of the complementary information between the different original images and eliminating redundant information, highlighting important information, and the information of the fused image are more complete and easier to understand than a single image. In practical applications, multi-source image fusion technology has been widely used in many fields, including the military field, medical research and diagnosis, industrial control, and remote sensing.Firstly, this thesis introduces the background of image fusion, the development status and practical applications, and elaborate on the three levels of the hierarchical classification and focuses on the multi-scale analysis that belongs to pixel-level image fusion. Secondly, this paper elaborates the Contourlet transform theory and nonsubsampled Contourlet transform theory (NSCT).NSCT is improved on the base of Contourlet transform and has a translational invariance, not including the image sampling operation. NSCT can not only extract the source image edge contours and other details, but also avoid the spectrum aliasing, because each child image with the same image size. By contrast to other multi-resolution transform methods we get the conclusion that NSCT has its own superiority as multi-scale decomposition and reconstruction tools in image fusion. The number of the NSCT decomposition layers and each scale decomposition direction is decided by decomposition parameter, thereby affecting the performance of image fusion. Because it can not be too small or too large, we need choose the best decomposition parameter for specific images.Multi-focus image is obtained by different imaging focal point. The multi-focus image fusion needs to extract clear areas of focus images through some fusion algorithm. For the characteristics of multi-focus image, an image fusion algorithm is proposed combined with the NSCT, and we can get a clearer image through the new proposed algorithm. The proposed fusion algorithm can reduce the impact of registration error. The infrared and visible light images reflect different aspects of the object, for the fusion of different images contains complementary information, we hope the new complete fused image can effectively inherit the spectral characteristics of the visible image and preserve the target information of infrared image. An image fusion algorithm is proposed combined with the NSCT, experiments show that this method can preserve the useful information of the image and contour edges of images are retained effectively. Finally, we have built the image fusion system demonstration platform through the graphical user interface GUI in MATLAB.