The Block Matching Algorithm Ds Video Judder

With the development of modern technology, especially the birth of large-scale computer, the use of computers to handle complex issues in real life to become a development trend. The design of image processing algorithm is the best combination of computer technology and image processing.The anti-shake technology has different research method at different times, each method has its own characteristics and application objects. The development of electronic equipment, technologies of computers and video image processing algorithm make the electronic anti-shake become the focus of research in the field of anti-shake. This paper makes the electronic anti-shake as the object of study. Meanwhile, studies of other mode of anti-shake has become mature, there is a lot of improvement space for the electronic anti-shake. We make research with it in system.The electronic anti-shake technology make full use of the characteristics of large-scale data processing in computer and the advantages of wealth graphic image processing algorithm. By researching the principle of image stabilization, we complete the anti-shake processing of the jittering video.Firstly, this paper studies both domestic and international classical algorithms about the anti-shake technology, and the latest research in this area recently. We analysis and summarizes the processing, the basic concepts, parameter model, evaluation criteria and so on, which involved in current treatment technology, then introduce the main technology in various processes. In which, we analyzes the various algorithm in the stage of motion estimation especially, from two aspects of accuracy and efficiency. We summarize the common defect of current motion estimation algorithm. Based on previous research, we finally choose the block-matching algorithm and analyze the matching manner used frequently in the block-matching algorithm. Relative to the motion estimation, the remaining stages such as image preprocessing, motion filter, motion compensation, lie in the auxiliary position, this paper also introduce the algorithm which involved in these stages. We do the further research on a variety of block-matching research algorithm. On the basis of analysis of various algorithms, DS algorithm eventually selected as the algorithm which used for searching target point.Then, in the gray-projection algorithm, we propose a new algorithm finding valley of cross-correlation curve called step self-adjustment search algorithm (SAS). Compared with the searching algorithm which used commonly in terms of the valley, the new algorithm we proposed has the trait of flexible adjustment of step length. It not only can move fast towards to target point but also can precise locate the target point. This, this algorithm prevent from searching the impossible area. We make experiment with the three point search (3PS) and step self-adjustment search (SAS), according to the analysis of experimental data, the new algorithm in the process of searching target point can reduce the total number of points which has to be traversed.Next, this paper proposed a new algorithm which combined gray-scale projection with block-matching algorithm. The algorithm uses gray-scale projection algorithm to perform the first matching and rough compensation, in which a new scheme of searching named step self-adjustment search (SAS) is used to decrease computation. Here, gray-scale projection plays a role of making assistance for the next steps. Then it uses block-matching algorithm for the second matching and precise compensation, which Diamond Search (DS) algorithm is employed for improving efficiency greatly. Therefore the algorithm is a coarse-to-fine process for image stabilization. The combination of Gray-scale projection and block matching algorithm can work very well. Experimental results show that the new algorithm well preserves high stabilization accuracy, at a dramatic reduction of computational complexity.Finally, the whole work of this paper and the future research are concluded and discussed.