Research On United Image Stabilization Based On Feature Tracking And Gray Projection

Nowadays, Electronic Image Stabilization (EIS) technology is an important research area of video processing and computer vision. It realizes the stabilization of image sequence via computer algorithm, so it has many good qualities such as real-time, high precision, low power and has been widely used in military affairs and civil cameras.Firstly, starting with the basic principle of EIS, this paper introduces two common used motion models. Then, it puts emphasis on common algorithms of global motion estimation including Bit Matching Algorithm(BMA), Gray Projection Algorithm(GPA), Features Tracking Algorithm(FTA) and Bit Plane Matching(BPM). Detailed analysis is made to describe the advantages and disadvantages of these algorithms.Secondly, this paper mainly studies Gray Projection Algorithm according to translational jitter estimation. And considering of practical application, the algorithm is improved for reducing the effect which due to amplitude of image edge in correlation calculation, Matlab is used to make image stabilization simulation test. Then, the paper puts emphasis on Hough Transform used for linear feature detection in image. A LUT Hough Transform based on second quantification method which can resolved the problem on large amount of calculation in traditional Hough Transform is presented to estimate rotation offset in images, makes image stabilization simulation test.Finally, a image stabilization algorithm based on improved Hough Transform and Gray Projection Algorithm is presented according to the video image sequence which includes complex jitter(translational jitter and rotation jitter). And this method is used to make EIS simulation test and the concepts of residual image and image sequence fidelity are introduced to evaluate the simulation test. The result shows that this new algorithm can estimate complex jitter in video image sequence, remove the jitter and improve the stability of video image sequence, performance stabile and accurately.