Research On Electronic Image Stabilizing Technology Based On The Horizon

Electronic image stabilizing technology is a hot topic technique in dynamic video field. This thesis designs an electronic image stabilization algorithm based on the horizon after working over the existing electronic image stabilization algorithms and practical application environment of the sea. First, we extract the horizon, and then, we use the location and the angle of the horizon to obtain motion compensation values, and finally, we make use of the affine transform and the bilinear interpolation to rectify the image sequence. Our research works include:1. Because of limitations of classic line extraction methods, a new algorithm of the horizon detection based on SVM is presented. The results of experiments show that this method enjoys its merits on robusticityand efficiency.2. For complexity and variability of the sea environment, the thesis discusses advantages and disadvantages between Hough Transform and SVM. Furthermore, a new algorithm for detecting horizon combining these two methods are presented. In this algorithm, a method of assessing the effectiveness of the horizon is proposed, which is used to switch two different methods automatically.3. Based on the horizon detection, the thesis further designs an algorithm for motion compensation. The bilinear interpolation is employed to deal with images after motion compensation, which can make stabilization accuracy meet the sub-pixel level.Combined all the above mentioned methods, the electronic image stabilization algorithm based on the horizon detection is developed. Extensive experiments show that this system can make image sequence under regular weather achieve sub-pixel accuracy of image stabilization, and meet the requirement of real-time.