The Research Of Digital Aerial Image Stabilization Algorithm

Video stability is an important index of video quality, which directly affects the visual effect of video. Due to the jitter of camera carrier platform, aerial video image is not stable. In order to achieve image stabilization for airborne platform, the paper deeply researched digital image stabilization algorithms. Based on application scene, digital aerial image stabilization algorithm is proposed, which provides new ideas for the research of image stabilization for airborne platform. This paper completed the following work:Firstly, the basic theory of digital image stabilization was researched. The paper reviewed the common methods of motion estimation and motion compensation, and summarized their applicable scene and performance. Feature matching is applied to aerial image motion estimation. There is no fixed reference image to compensate jitter for dynamic platform, so the method of motion smoothing is adopted to compensate motion. This paper also described the evaluation methods of the digital image stabilization quality.Then, when traditional feature matching is applied to motion estimation, there exits degradation problem of feature distribution, which leads to decreased accuracy of motion estimation. To solve the degradation, sub-block feature detection was presented. Sub-block control can make feature points even distribution, reducing the estimation error. In order to improve the low efficiency of feature matching, the method of feature matching then tracking was proposed to estimate global motion from coarse to fine. It can enhance the accuracy of motion estimation effectively.Finally, motion compensation algorithms were deeply researched. There are some disadvantages to compensate motion by traditional motion filtering. According to the characteristics of aerial image jitter, this paper adopted the latest Hilbert-Huang transform for image stabilization. From non-linear and non-stationary signal decomposition, which is different from motion filtering, Hilbert-Huang transform can accurately separate the camera scanning motion and random jitter from global motion, which can effectively compensate camera jitter and improve quality of image stabilization.The proposed methods of aerial image stabilization were tested on different video sequences. The results show that the proposed methods have good accuracy and validity.