Research On Electronic Image Stabilization Algorithm Under Compound Motion Condition

Electronic image stabilization(EIS)is one of the popular technology on image and video processing.It can effectively solve the image jitter caused by the random motion of camera carriers for obtaining stable video images.Meanwhile,it can also improve the accuracy of subsequent image processing and the comfort of visual experience.Therefore,EIS has bright application prospect in the fields of military and civilian.Based on the research of EIS algorithm in compound motion condition,three key technical modules named motion estimation,motion filtering and motion compensation are deeply analyzed.Aiming at the low algorithm accuracy and processing speed caused by compound motion of image jitter,the works of this thesis are as follows:(1)In order to solve the problem of slow speed and uneven distribution of traditional feature point extraction algorithm due to the diversity of motion state and the variability of scene.The ORB feature point algorithm with adaptive threshold is applied in this thesis.Since the useful information in a jitter image is mostly concentrated in the center of the image region,the extraction speed of feature points can be effectively improved by dividing the interesting regions of the image.However,the image information differs in different practical scenes.If the feature points are extracted by the same threshold,the number of feature points will be unevenly distributed.So,combining blocked area of interest with the adaptive threshold algorithm could effectively solve the problem.(2)To improve the accuracy and speed of motion estimation,a method of optimum search area is proposed.The feature points of reference frame are taken as the center points,and the optimal search region is constructed by using the magnitude of the motion vector in the previous frame.It can effectively reduce the range of search area of LSH feature point searching and matching algorithm based on Hamming distance,thus improve the speed of feature point searching and matching.In addition,considering the low accuracy of the RANSAC algorithm,the PROSAC algorithm is adopted to eliminate the mismatched feature points and further improve the matching accuracy of feature points and the accuracy of motion estimation.(3)In order to enhance the image display after stabilizing,this thesis adopts the method of motion compensation algorithm based on the Kalman filter algorithm.The Kalman filter algorithm is used to obtain the random jitter component and keep the intentional scanning motion of the camera.The corresponding motion compensated image retains the real scene transformation in the front of removing the jitter.Meanwhile,the black area is eliminated due to the bilinear interpolation.In addition,an adaptive reference frame update strategy is proposed to make the inter-frame over-natural and display stable.(4)An experiment is carried out to verify the rationality and accuracy of the algorithm proposed in this thesis.From the subjective and objective analysis of the experimental results,it can be seen that the algorithm can effectively solve the problem of video jitter under compound motion,and the speed is 1.26 times faster than the contrast algorithm,which will achieve a good combination of image stabilization accuracy and real-time performance.