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

The first generation Curvelet transform has some shortcomings, such as it is so complex and also brought a huge amount of data redundancy.For the above shortcomings,it leads to the second-generation Curvelet transform,then introduces its principle, also puts forward two kinds of methods. One is USFFT algorithm and the other is wrapping algorithm. But taking the complexity of the problem into account, in my thesis, the wrapping algorithm is used.Original fusion will directly lead to the failure of the fusion processing. By comparison, we got that the infrared images have multiplicative noises, Low-contrast, and low-resolution. The denoising and the enhancement of the infrared images are proposed.The current image denoising algorithms have some shortcomings, such as the hard threshold algorithm, the edge of the image will have many "artificial" noise points. And the soft-threshold algorithm caused a amount of high-frequency information loss and the edge of the results are blurred. For the above shortcomings.Using this algorithm, we can not only enable energy adaptive noise suppression,but also can save the image edge information very well.Based on the short of the Curvelet transform in the express of the image detail, the algorithm based on the Curvelet transform and wavelet transform is proposed. The purpose of combining the two transformations is to take their respective advantages and weaknesses complement each other. so the long-edge information and point details of the fused image are well retained. At the same time, it also proposes a new coefficients selection method, a large number of experimental data show that fusion results can retain target information and spectral information better.Finally, the existing single-threshold and dual-threshold enhancement algorithm have their shortcomings. Such as the single-threshold enhancement algorithm may also enhance the noise. Dual-threshold enhancement algorithm may cause excessive noise suppression, that will cause that the details have been fuzzied.To solve the above shortcomings, it poposed the adaptive-threshold enhancement algorithm. A large number of experiments show that the algorithm is much better than the traditional histogram equalization algorithms, the single threshold enhancement and the dual-threshold enhancement.