Infrared Image Stitching Technology

As an method of opto-electronic imaging, infrared thermal imaging technology plays an important role in military detection and civilian technology field. Infrared image mosaic is a research topic in early warning and circle viewing system. Also, it provides a solution to the low resolution problem in the current thermal imaging.Based on the characteristics of infrared images, the thesis mainly discusses image registration which is the key part in image mosaic. The main purpose of the study is to effectively mosaic infrared images with an appropriate image mosaic algorithm.In the registration algorithm based on gray level, an improved SSDA algorithm with a combination of adaptive threshold and wavelet pyramid is studied. Because of the optimal search strategy, its speed could reach real-time requirements. In the FFT-based image registration algorithm, log-polar coordinator is introduced into the phase correlation to expand image registration from translations to affine transformations.In the feature point based image registration, the common corner detection algorithms are summarized. The principle, advantage and disadvantage of each algorithm are analyzed. An improved Harris algorithm is proposed. These corner detectors are compared from cost time, corner detection rate and repeatability rate through the simulation. Results show that the improved Harris corner detector is a robust method. But, all above corner detection algorithms are subject to image scale change.On the basis of the above study, combined with the characteristics of infrared images. An infrared image mosaic based on SIFT algorithm is simulated. In this method, the feature extracted and described by SIFT and the approximate-nearest neighbors of each feature are found out by BBF. Then the homography is calculated by a RAN SAC with filter. An improved get in and get out method is presented for image fusion. The simulation results show that the method can be applied for infrared image mosaic.