Digital Video Image Stabilization System Based On Block Matching

About80%of the information Human obtained is from their own visual system. Visual digital media technology has broken physical limits of real life, making the use of digital media information effectively, in protable applications, there can be some motion in image collection system, making images blurred. Hence a digital video image stabilization is needed to improve its quality.According to the difference of hardware equipment, technical means and working principle, image stabilization technology is divided into three categories:mechanical image stabilization technology, optical image stabilization technology and electronic image stabilization technology, a digital video image stabilization, aided by computer, takes advantage of image processing algorithms, in motion estimation of dynamic video image frame and determination of random motion for a more stable video series. This technology has the merit of high accuracy and speed. The main purpose of this thesis is to study and to achieve the digital video image stabilization system, focus on the block matching algorithm in motion estimation module, gray projection algorithm and feature corner algorithm and put forward a new improved algorithm based on full search block matching method. In motion compensation module, Kalman filtering, curve fitting, low-pass filtering are studied, and finally real-time Kalman filtering is selected as the motion compensation module to reduce the shake of the video, and then a self-adaptive Kalman filter is put forward on this basis.Three basic and an improved digital video stabilization system are implemented by using VC++, and then tested by three videos in different three movement mode. From the sense of timeliness, the results show that gray projection algorithm is the fastest, followed by block matching algorithm, while feature corner algorithm is the last one. The new algorithm based on block matching can reduce the calculation time by15%-30%. In addition, the improved self-adaptive motion compensation module get a better motion tracking compared compared with the original Kalman filter.