Study On The Techniques Of Digital Video Stabilization Based On SIFT Feature Match

In recent years, handheld camera equipment and airborne camera equipment used more and more widely used, when these cameras carrier movement, camera equipment will be a corresponding movement, resulting in the shooting of the video will also have the corresponding jitter, affect people’s visual effects, reduce the visual quality of the video, video jitter can cause tracking system or target recognition system false alarm or missed alarm in the military field. Digital video image stabilization in digital image processing method to eliminate the jitter of the video, so as to produce stable video sequence.Reading on the basis of a large number of domestic and foreign literature, the first classic digital video stabilization algorithm summary analysis of its pros and cons and usage, and on this basis, for the same time there is a rotation, pan and zoom complex jitter, image stabilization algorithm based on SIFT feature extraction. In this thesis, the research aspects:(1) In this paper, the classic image stabilization algorithm are summarized and compared the classical algorithm performance and application shaking. Moving different mathematical models for different jitter the movement filtering algorithm in image stabilization, this paper uses the Kalman filter to solve the problem of the presence of active scanning with shot image stabilization.(2) In this paper, different feature extraction algorithm are summarized, including the Harris corner extraction, SIFT feature extraction. And SUSAN corner extraction algorithm, through actual shooting video simulation, these types of feature extraction algorithm, extraction accuracy, the number of feature point extraction and matching time to compare the performance of several algorithms, and the final choice SIFT algorithm as a characteristic point.(3) The presence of translation, rotation and scale changes at the same time the complexity of jitter, extracted SIFT special. Intrinsic point image stabilization system, first on the existence of a complex and shaky video, SIFT feature point extraction and matching, matching feature points RANSAC algorithm mismatched points removed, and then use the Kalman filter active Scan component to extract the jitter components, the video image stabilization.For the complex jitter actual video, this paper carried out a large number of simulation experiments, tested the algorithm accuracy and real-time, with classic stable like algorithm are compared, the experimental results show that the presence of moving objects and complex jitter image, the proposed algorithm can achieve good results.