Study On Fast Electronic Image Stabilization Algorithm Using Feature Registration

With the rapid development of computer science and IT technology, image andvideo processing technology are widely applied, and its performance requirements arealso rising. With the widespread popularity of the rapid development of IT technologyand portable mobile electronic products, images and video has become an importantinformation media, to improve image quality requirements. With its advantages oflow-cost, high-precision of Electronic image stabilization technology, embedded andstability by more and more attention. It has a very important theoretical significance andpractical value.The papers on the basis of analysis of the merits of traditional electronic imagestabilization methods, with a focus on feature-based matching method and its discussion,for the time-consuming long operations shortcomings, propose a contour sparsedescription image stabilization algorithm. Guarantee the accuracy significantly reducethe computational complexity of the algorithm has a certain meaning and applicationprospects. The major research in this paper can be summarized as follows:①Research of image stabilization technology is surveyed, the classic imagestabilization methods is recalled. The basic processes of the electronic imagestabilization are introduced. The main methods and description of its key priorities isrecalled. Common classical algorithm in electronic image stabilization, and analysis ofthe advantages and disadvantages of each algorithm are also are Introduced. Theresearch is focus on feature-based matching method and relative to the advantage ofother algorithms as well as the shortcomings of it. The main work in this paper is theexpanded feature-based approach, the above work paving the way for the work below asolid theoretical foundation.The research progress at home and abroad and the common methods of EIS arereviewed and summarized. In this thesis, methods of the electronic image stabilizationare described; the key technology of the realization is discussed. Several typicalalgorithms of EIS and their advantages and shortcomings are introduced and analyzed.It emphatically points out that the algorithm based on feature matching is considered asthe important method, because of its perfect performance to deal with the complexdither between image frames. The main work in this paper is the expandedfeature-based approach, the above work paving the way for the work below a solid theoretical foundation②The feature matching method has high accuracy but also brings a tremendousamount of computing, but how to balance between precision and real-time is also alarge number of heavy and difficult problem of the motion estimation algorithm. In thispaper, to solve the problem of large amount of computation for feature matching, a fastelectronic image stabilization method is proposed, which is based on contour sparsedescription. Firstly, well affine invariant image edge is got, after the edge expansion toget the corner-intensive areas and extract the corner features. Then getting thedistribution of relation uniform angular corner features and matching. Finally, we getthe global motion estimation parameter.③For the existence of the self-scanning and random jitter of the video sequence, itmust be effectively isolated from the autonomous scanning of the camcorder. This paperis based on the basic principle of the motion filter, combined with the spatial position ofthe image model, and the damping coefficient filter is introduced effectively smoothingseparated random jitter and the self-scan of the camcorder.④For the black side effect after the images direct compensation, the analysis of theprinciples and characteristics of the image cropping, unresolved problems crop broughtthe introduction of this thesis, image stitching technology undefined region to fillrefactoring. The method is effective to improve the visual effect of the output imagesequences, to avoid image information loss and the decline in the quality of thetraditional methods.