Research On Electronic Image Stabilization Algorithm Based On Phase Correlation Method

Due to the carrier shaking of airborne, shipborne, on-board camera equipment at work, the captured video often tends to jitter. The jitter will seriously affect people’s visual observation, and is not conducive to subsequent various image processing algorithms. Electronic image stabilization method using digital image processing techniques to achieve the stability of the video sequence, because of its small size, high accuracy and low power consumption, has received wide attention of scholars.Firstly, the research status of image stabilization technology and the current main technical problem is introduced. According to the execution order, the basic principle of elaborated motion estimation, motion filter and motion compensation module are described in detail. And several mainstream algorithms of each module are analyzed and simulated to compare their advantages and disadvantages. Then on the basis of traditional phase correlation method, we give a method to determine the rotation and scaling factor, and put forward a method of determining the subpixel accuracy based on the interest area, this method can effectively improve the efficiency without sacrificing accuracy. Then again, in order to improve the computational efficiency of the algorithm, we introduce three algorithms optimization:ROI-DFT transform, FFTW library, image partition parallel processing; the latter two methods are applied to the final hardware implementation, which significantly improves the efficiency. Combined with previous work and exploration of the three possible impacts on the robustness of the problem:random noise, aliasing, spectral leakage, a solution is put forward. In experimental section, designed algorithms are simulated in Matlab, performance analysis from two aspects of difference image and continuous frame PSNR show that the designed algorithm has a better recognition on translation, rotation and scaling transformation; and the accuracy is superior to other algorithms. Finally, comply the proposed algorithm on the DaVinci image processing hardware platform, the result of implementation time is 29 frames/sec.