Research On Fusion Algorithm Of Color Visible Image And Infrared Image

Image fusion aims at integrating multiple images obtained from different sensors to a single image. Such an image is more accurate and reliable and it contains abundant information. Image fusion technique has been widely applied in many fields, such as military, medical care and artificial intelligence. Image fusion has a broad prospect.Color visible image can display the details of a scene very clearly, but it is vulnerable to light conditions. In contrast, infrared image has unique advantages in detecting heat-emitting objects, but it is poor in the quality. Fusion of the two images can make up their shortcomings and will help to show the objects. What is more, it has richer scene detail and better visual feelings. The dissertation does some research on this issue and discusses the criterions of the fusioned image. The following works are included:Firstly, some research is done on the fusion algorithm base on pyramid decomposition theories. Laplacian pyramid, contrast-pyramid, gradient-pyramid and morphological-pyramid are included here.Some simulations are also tried out to test these algorithm.Secondly, several criterions for the fused images are discussed, which include cross-entropy; mutual information and edge-appraisal. The assessment of fused images is obtained by these criterions.Then, we combine Wavelet-based Contourlet Transform and lab color space together to get a new fusion method. The characteristic of infrared image and color visible image is also taken into account to accomplish the fusion of the low-band coefficients. Simulations are done to illustrate the priorities of this algorithm.Finally, pulse-couple neural networks (PCNN) are mentioned to be combined with non-subsampled contourlet transform (NSCT). The principles and its models are introduced. We propose an effective way of image fusion based on them and carry out simulations about this algorithm.