Research On Fusion Of Infrared And Visible Images Based On The Tetrolet Transform

The information of one single image can’t meet people’s demand with the rapid development of the science and technology, the improvement of technology of various sensors, and the improvement of people’s living standards requirements. Furthermore, in order to detect the target for the soldiers to leave sufficient time to prepare in the military battlefield, so the image fusion technology becomes more and important, and the research on this technology have great significance. The main research of this thesis is the fusion between the infrared images and visible images which are one pair of complementary sensor image, and what’s more the information of the fusion image is more abundant.We make some analysis and study to some commonly used algorithms, make comparison between these algorithms, and discuss the advantages and disadvantages of these algorithms. An improved tetrolet transform is used in the fusion of infrared image and visible image which was proposed according to the target object of infrared image is high-energy and the detail information of visible image is rich. First of all, this algorithm is improved in the selection of tetrolet transformation template, no longer to choice the minimum high-frequency coefficients, but the high-frequency coefficients of the first order maximum norm as standard. This is to retain more original high frequency details in the high frequency region, improve the sharpness and amount of information of the fusion image effectively. Secondly, remove the hard threshold function with a loss in terms of de-noising, but it can effectively preserve the fine details, making the inverse transform to improve the quality of the fusion image. The contrast of the fusion image improved through increasing a process which is the choice of template with peak signal-to-noise ratio PSNR as the selection criteria in the inverse transformation. The final step is to improve the fusion rules by changing the self-adaptive local area energy fusion rule. We increase a factor which is energy proportion, according to the energy proportion to zoom the weight appropriately. This can effectively preserve the target feature information of infrared image and the background information of visible image.In this thesis, some experiments are made by selecting two different background images to compare. One set is the city’s night scene with abundant people target information. Another set is the image of field environment with complex background information. We make experiments by using MATLAB simulation software and the results show that the proposed algorithm is far better than some other commonly used image fusion algorithms. In the subjective evaluation, the target characteristics of the fusion image obtained in this paper’s algorithm are better, and the background information of visible image is also abundant. In the objective experimental data, the average gradient value AvG, information entropy IE and peak signal to noise ration PSNR improved greatly, which effectively verified the validity and feasibility of the proposed method through the objective data. In the implementation of the program, the tetrolet transform template and inverse transform template are both represented by one large matrix through using MATLAB’s powerful data processing capabilities. In the calculation process, calculate with matrix multiplication instead of cycle counting which not only solves the problem of selecting a template, but also saves much time and operation costs.