Research Of Digital Image Stabilization Method On Airborne Video

Video image plays an importantrole in people’s lifeas an important information media, while airborne video occupies important position in security, military and other fields. As the spread of smart city and the mobile internet, video monitoring will has been further development in the smart home and smart car, it also put forward higher requirements about the quality and real timeof the video. It is inevitable to be bumpy in the process of vehicle movement, at this time, the video acquired from vehicle traveling data recorder video will have the phenomenon of random jitter. In order to eliminate the random jitter in the video sequence, the purpose of this paper is to research the method of digital image stability of airborne video.This paper introduces the process of development and the research statusof digital image stability technology at first, the method and the composition of DIS(Digital Image Stabilization) will be introduced then, it also analysis the model of video-filter, the basic principle and method of motion estimation and motion compensation.For the widely used DIS algorithm, sub-block matching, this paper have a research on four kinds of search strategies, full search,three step search,diamond search and adaptive rood pattern search, and carried on the comprehensive evaluation according to the search speed and quality of video. Because of the real-time performance of sub-block matching is not good enough at the situation of great deviation between video sequence, this paper has a further study of gray projection algorithm, and has made improvement at 3 points about traditional gray projection algorithm: Firstly, enhancing gray-scale contrast through image preprocessing, this solved the problem of gray-level changed not obvious; Secondly, putting forward sub-domain vector separation algorithm, this will get rid of the influence of moving targets and make the global movement estimate more accurate; Thirdly, proposing adaptive one-way search strategy, this strategy will have different step length after deciding the search direction by the way of analysis the mathematical properties of related curve, it make a great improvement comparing with full search, and this is really useful to airborne video about the requirements of real time. This paper had also proved these improvements is effective by experimental simulation.In the end of this paper, the method of motion compensation were analyzed.In order to keep the part of common movement on airborne video, this paper had aresearch on the least squares curve fitting method, average filtering method and kalman filter method, due to kalman filter just need to know the posteriori estimates and a posteriori estimation of covariancein a moment, this is why it has a good performance on operation, and is fit to DIS which has a high requirement on real time.Therefore, this paper uses the kalman filter as the global motion vector filtering method.