Electronic Image Stabilization System Based On Delaunay Triangulation For Implementing Motion Estimation

With the rapid development of the technology of video surveillance, electronic monitoringexperience from early indoor monitoring to the current in the new motion imaging platformssuch as electronic monitoring equipment placed on board vehicles, boats. The effect of thisnew camera system affected by the change of posture of the carrier itself, monitoring thevague and shaking of the image sequence, so it brings the development of image stabilizationtechnology. Since mechanical image stabilization and optical image stabilization need to use aspecial device to determine the movement offset, electronic image stabilization brings us newideas, it can determine the offset by the image offset, and then making the image framemotion compensation. Electronic image stabilization is more flexible, smaller consumption. Itis gradually developing as image stabilization technology trends. Electronic ImageStabilization applies to border posts, reconnaissance vehicles, shipboard, indoor and outdoormonitoring and so many other application areas, and a lot of researches about it have been putinto application.The core technology of the electronic image stabilization is how to image matching.Efficient image matching strategy should take into account real-time and estimate accuratemotion vectors between frames. So the key to solve the problem of image stabilization is howto accurate motion vectors according to the image sequences. After image motioncompensation, and the system can output the output stable image sequences.This paper describes the current popular electronic image stabilization algorithms: block-matching method, the gray projection algorithm, bit plane matching algorithm, featurematching algorithm and so on. The advantages and disadvantages of these algorithms in theaspects of the image stabilization are to be analyzed. The paper analyzes the problems whichthe image stabilization technology should consider. The problems include the calculation ofthe motion vectors, the determination of the parameter model, and the moving way ininterframe. The algorithm in this paper combines the advantages of the gray projectionalgorithm in translational motion areas, it also combines the Harris detector corner, Delaunaytriangulation characteristic triangle, the thought of scale invariant feature transform (SIFT features) to calculate the characteristic triangle descriptor. Finally, it uses optimal matchtriangle to calculate the affine transform matrix, and eventually makes the video imagestabilization.In summary, this paper presents an electronic image stabilization system based onDelaunay triangulation for motion estimation. Firstly, the current frame achieves the initialtranslational compensation relative to the reference frame by using gray projection (GP)algorithm. Secondly, we use Harris corner detection algorithm to get the corners of the currentframe and the reference frame. We can get a Delaunay triangulation for these corners pointsby using divide-and-conquer method. Each vertex of the triangles are feature corner points,then we calculate characteristics triangle descriptor. We can calculate the affinetransformation matrix after we get the optimal matching triangle. Finally, the algorithm canmake the video image stabilization by using the affine transformation matrix.In the experiments, the experimental video uses only Delaunay characteristic trianglematch to get image stabilization, and uses only the gray projection algorithm to get imagestabilization, and the gray projection algorithm fusion with Delaunay characteristic triangularmatch algorithm to get mage stabilization. We have evaluated the proposed method ondifferent video sequences. The experimental results of each step in the algorithm will bedescribed and analyzed. The experimental results show that this method can be real-timeimage stabilization.This paper makes a meaningful exploration in how to make the video image stabilizationhigh-precision. There are a wide variety of electronic image stabilization technology and howto make electronic image stabilization technology faster and more accurate precision willcontinue to be a challenging issue.