Research Of Electronic Image Stabilization Algorithm For Video Surveillance

Electronic image stabilization technology Electronic Image Stabilization (EIS) is a research focus in video processing, computer vision and other related areas. It is widely used in military and civilian, to improve the stability of the camera motion vector and clarity. With the rapid development of large scale integrated circuits, electronic image stabilization applications are increasingly widespread. Electronic image stabilization makes up the disadvantages of optical image stabilization, electrical and mechanical methods, and as electronic image stabilization, the algorithm is particularly important. The EIS is very meaningful for all occasions target tracking and detection. Real-time electronic stability such that the electronic image stabilization methods in a timely manner to stabilize the image sequence, camera system and tracking system allows the observer will not be the interference of camera itself movement, to achieve real-time video tracking and image stabilization purposes.Based on the study of the electronic image stabilization technology, this thesis analyzes the basic principle of the mathematical model of electronic image stabilization, the image stabilization technology, the main features of electronic image stabilization, and the main steps. This article analysis of several types of common electronic image stabilization algorithms-- representative point matching (RPM), block matching (BMA), projection algorithm (PA) and the bit plane matching (BMP). By using PSNR as the evaluation criteria for algorithm, this thesis makes the evaluation of the accuracy and the speed of the algorithms. The results reveal that the PA could maintain high accuracy while the mathematic works is not heavy, and the PA is suitable for electronic image stabilization system with high accuracy and high real-time.This vibrate of the surveillance video is mainly due to small-scale translational motion, which is caused by environment factors. The EIS for surveillance video should meet the accuracy requirements while also ensuring real-time property. Thus, for the accuracy demand of image stabilization algorithm, made the image preprocessing to enhance image contrast. At last, based on PA, this thesis reveals a fast search algorithm– three point method. Because of the correlation curve based on PA has the single peak characteristic, we can determine the relevant values'reducing trend while searching process. In consequence, while maintaining the accuracy of searching process, this algorithm'motion estimation faster than the classical full search block matching was five times higher. The experimental results show that the three-point method based on PA is able to meet the requirements of the EIS for surveillance video in the accuracy and speed.