Research On The Algorithm Of Real-time Digital Image Stabilization And Motion Object Recognition

The digital image real-time stabilization is a process which can make compensation between the video sequences for camera equipment and make the output image clear and steady by reducing or eliminating the instability. The image stabilization is widely used in hand-held camera, mobile photography equipment and has a very important influence on the post processing of video, such as television surveillance system, moving object tracking system, and navigation system. The block matching method is a common processing algorithm. The image is divided into equal-sized image sub-block to match, and the algorithm is simple, stable operation and running fast.In this paper, an improved block matching algorithm is proposed by analyzing the basic principles of block matching of digital image stabilization and other common algorithm. In order to make the matching efficient higher, the algorithm takes the advantage of the feature information of the image. In order to make the system to adapt to the different environment, Euclidean Distance is used as the matching principle. A classification tracking algorithm is designed for the application of the algorithm of target recognition, and a multi-target classification and tracking process is realized.The results show the performance and stability of the improved matching stabilization algorithm is improved compared with the traditional block-matching algorithm. Meantime, the improved algorithm is simple in principle and easy to implement.Since the stabilization algorithm processes the unstable video image, the algorithm should adapt to more complex and different situations. In subsequent work, the algorithm will be optimized and improved on the rotation of the video image, the change of the focal and smart jitter judgment. The performance and effect of the algorithm should be improved in a more complex environment.