Improvement Of Motion Estimation Algorithm And Application Of Electronic Image Stabilization Technology

Image stabilization technology is designed to eliminate the impact of the camera device carrier jitter on the shot video, including mechanical image stabilization, optical image stabilization and electronic image stabilization. Compared to mechanical image stabilization and optical image stabilization technology, electronic image stabilization technology directly uses digital image processing technology to determine the image pixel offset between frames and then compensate it to achieve stabilization. Therefore, the device of it has advantages of small volume, low cost, applicable range widely and electronic image stabilization has become a hot research field of video processing and computer vision in recent years. The accuracy and real-time performance of motion estimation is the key technology of electronic image stabilization, and also the challenge of it. This paper studied electronic image stabilization algorithm based on feature points matching and its main research contents and results are as follows:Firstly, a motion estimation algorithm based on the combination of feature points matching and histogram statistics is proposed. For video sequences, the proposed algorithm firstly arrange sub images in the reference image and the current image and then does feature point extraction, description and matching in the corresponding sub image. Continuously, those error matches are eliminated by the principle of motion consistency motion and these accurate matches are obtained by the histogram statistics. Lastly, the affine transformation model parameters which stand for the global motion vector are calculated by using the average position of accurate matches.Secondly, the proposed motion estimation algorithm combined with motion filtering to achieve electronic image stabilization, and compared it with the domestic and foreign advanced algorithms after programming the entire electronic image stabilization process. The global motion vector includes the intention motion vector and the random jitter vector, and the k frame latter Gaussian filter is used to filter out the random jitter vector. Then the current image can be transformed into a stable image to form a stable video sequence by using the affine transformation model which contains the intention motion vector. The quality and accuracy of the proposed algorithm are validated through computing efficiency analysis and objective quality of image stabilization.Finally, the proposed algorithm is applied to two kinds of practical problems. On the one hand, it has been used in multimedia mobile phone alarm and remote handle emergencies based on mobile APP system. The video sequence which is shot by intelligent mobile terminals and has been stabilized by the proposed algorithm not only meeting department officers viewed clearly and accurately determine the alarm situation, but also assisting them to carry out accurate analysis of cases. On the other hand, after processing the random jitter in aerial traffic video sequence, it can improve the accuracy of moving vehicle detection which laid an important foundation for the subsequent vehicle tracking and traffic parameter achieving.