Surveillance Video-Oriented Research Of Electronic Image Stabilization Technology

Compared to the traditional mechanical and optical image stabilization case, Electronic image stabilization is easy-to-use and has small size, light weight, low power consumption, high precision. It will make the stabilization equipment small and portable. It has a broad application ,in the domain of military and civilian. With the maturity of theoretical study and the appearance of image stabilization products, electronic image stabilization will be one of the development direction of modern image stabilization technology.Surveillance Video-Oriented Research of Electronic Image Stabilization Technology is studied in this paper, and the goal is to implement the image stabilization algorithm on the DSP. Based0 on the analysis of surveillance video features, a integrated electronic image stabilization system is designed based on hardware and software conditions. The research paper focuses on the electronic image stabilization algorithm in the system, realized and improved two electronic image stabilization algorithm. For the fixed camera translation jitter, the projection estimate and direct motion compensation were used to remove the jitter; For the movement camera translational jitter, projection estimate and kalman filtering were used. When rotating movement exists in the surveillance video, the SIFT proposed algorithm and rigid model were adopted to get the rotation angle, then compensating rotating movement, making the surveillance video only have translation jitter. Also, after the implement of two traditional compensation algorithm, a local mosaic and image inpainting based compensation algorithm is proposed to recover the blank region that emerged at the edge of images in the surveillance video after the motion compensation. And the algorithm works well.The proposed algorithm is verified through testing video and real video. The experimental results show that the study of this paper helped to improve the stability of the video, the getting of the rotation angle and the compensation of the undefined blank regions very well. And the proposed algorithm can meet the requirements of electronic image stabilization.