With the video capture device widely used in various fields, the stability of image information is becoming more and more important. Electronic image stabilization is a new method to stabilize image sequence, which uses computer image processing technology to achieve the goal. It has a broad application prospects in both the military field and the civilian field.The carrier in the low altitude with high speed, such as artillery shells, missiles and so on, have to maintain its own self-rotation in order to maintain its balance. Because of its own spin, the image sequence captured by the video capture device in the carriers is unstable. The object of this research topic is to eliminate the rotating motion vector generated by self-rotation, and make the image sequence to be stable.This paper first describes the basic principles, common methods and processes of electronic image stabilization. After analyzing the electronic image stabilization algorithm has been used, three kinds of rotating electronic image stabilization methods are designed according to characteristics and the movement model of the rotating image sequence. These three methods use algorithms based on transformation model, SIFT feature matching and circular projection to estimate rotating motion vector. We simulate and verify algorithms with MATLAB and OpenCV, then used MSE, PNSR and sub-image to evaluate stabilization accuracy of these algorithms. The algorithm based on SIFT feature matching and the algorithm based on circular projection have a higher accuracy of image stabilization, the algorithm based on transformation model is somewhat less effective. Algorithm based on circular projection has a faster processing speed, algorithm based on SIFT feature matching and algorithm based on transformation model is much slower.At last we choose the method based on circular projection to be realized in the embedded platform according to the stabilization accuracy and processing speed. This paper designs and realizes an electronic image stabilization system based on Davinci 6446 embedded platform, which uses ARM + DSP dual-core processor chip as the center, and uses codec engine to communicate with dual cores. This paper discusses three ARM-side modules: video capture module, master procedures module and video display output module, one DSP-side codec engine implementation module of rotating electronic image stabilization algorithm. Experimental results show that the rotating electronic image stabilization system has a high accuracy of image stabilization, it can reach a processing speed of 38.3/s for video with resolution of 480×320, which meets real-time requirements. |