Research On Fusion Method Of Infrared And Visible Images In Super-wavelet Domain

Image is an important information source of human perception. With the increase of the image acquisition mode, the requirements of image fusion in the modern image processing system are also improved. Different source Images have different fusion requirements and fusion purposes, and the fusion algorithm is different either. Infrared and visible image fusion technology can extract features of different band images, so that the fusion image can get better target characteristics and clear scene information.In this thesis, we mainly study the infrared and visible image fusion algorithm based on super-wavelet domain. This thesis mainly discusses the main ideas and the realization of the two kinds of super-wavelet transform. One of them is Tetrolet transform, which is a new super-wavelet transform, and the other is a maturity super-wavelet transform NSCT. There are two improved fusion algorithms: A new algorithm based on improved Tetrolet transform and Pulse-Coupled Neural Network and A new algorithm based on self-similarity in NSCT domain. The main research work of this thesis includes:1. In order to overcome the problem of the Blocking Effects, a new algorithm of infrared and visible image fusion based on improved Tetrolet transform is proposed. Firstly, the two source images are decomposed into low and high frequency subbands by Tetrolet transform. Then, the low frequency subband coefficients fused based on an adaptive regional energy weighting; while for the high frequency subband coefficients, based on improved PCNN. Finally, the fusion results are reconstructed by Tetrolet inverse transform. The experimental results show that, the fusion image of the proposed algorithm can obtain a better target characteristic and has better subjective visual effect and objective quantitative index. And this algorithm is not only suitable for infrared and visible image fusion, but also has a good effect in other multi sensor image fusion.2. NSCT is an improved Contourlet transform, which not only has the translation invariance but also has sufficient redundancy information, and it can be more effective to extract the information from the source image, so that the fused image is more consistent with the human visual characteristics. Then the image fusion algorithm based on improved region energy and self-similarity in NSCT domain is proposed. Firstly, the two source images were decomposed into low-frequency subbands and high-frequency subbands by NSCT. Then, for the low frequency subband coefficients, an adaptive regional energy weighting image fusion algorithm was presented; while for the high-frequency subband coefficients, based on self-similarity. Finally, the fusion images were reconstructed by inverse NSCT. Compared with some existing algorithms, the algorithm of infrared and visible image fusion results in the human eye subjective visual standards and objective evaluation criteria have obvious advantages, and in other multi sensor image fusion field also showed an excellent performance.