Research And Implementation Of Electronic Image Stabilization Technology

With the progress of science and technology,as a convenient and fast method electronic image stabilization has been widely used.The camera equipment is affected by environmental factors,which can cause shake,causing instability and blurring of video or images.This not only seriously affects people to obtain useful information from video images,but also causes great inconvenience to subsequent processing and utilization of video images.Electronic image stabilization eliminates shake by directly processing the acquired video image.This article focuses on the rotation and translation jitter in infrared video,and proposes a comprehensive method based on gyroscope and feature point matching.This article elaborates the basic theory of image stabilization in detail,introduces the mechanism of video blur elaboration,image motion model,and constructs a mathematical model of electronic image stabilization.Several commonly used motion estimation algorithms are discussed in detail,then analyzes their benefits and drawbacks,and make some improvements.Aiming at the problem that traditional algorithms are not sensitive to rotational shake in video,this article uses a gyroscope to record the rotational angular velocity of the camera equipment during video acquisition.Perform Kalman filtering on the recorded data.After linear interpolation,integrate to obtain the rotation matrix,and then obtain the accurate camera rotation operation vector.This method can effectively avoid the disadvantage that the gray level of the infrared image is not obvious,and reduce the amount of calculation in motion estimation.Aiming at the problem that the gyroscope cannot obtain the translation shake of the camera equipment,this article uses SIFT feature point matching to perform translation motion estimation.The algorithm firstly finds the feature region by Sobel operator,after that match the feature points on the feature region.The RANSAC algorithm filters the mismatches,and through the improved k-neighborhood,it avoids human intervention to set a threshold and improves the applicability of the algorithm to different scenarios.For the method of the paper,collecting some video perform verification.Experimental results indicate that the method in this paper can effectively remove shake in video.