Multi-video And Multi-image Real-time Registration

Image registration technology can be used to many fields, such as military affairs, medical science, remote sensing and computer vision. Because Real-time registration of video can be used for guidance of telemedicine in operating room, specialists can observe situation of operating room in real-time wide field of vision and give telemedicine guidance when they do not enter into operating room. This paper discusses a kind of research approach which can realize multi-video and multi-image real-time registration and proposes three acceleration methods to increase the speed of video registration.The first, the process is divided into two periods, initialization period and real-time period. In initialization period, we utilize improved SURF to extract interest points in the first image. In real-time period, we realize the images’registration. The experiment shows that it can void double counting and reduce the computational work using the division of registration stages. The second, the valuable interest points are in interest areas of images which can be select by region selection algorithm. The third, the interest points that have stronger robust locate in the nearby of image edges. SURF algorithm is improved by Canny algorithm which can detect the edges.The experiment shows that it increases the speed of interest point extraction using Canny-SURF. The three methods reduce the computation availably and cut down the time of video registration in the case of the same registration accuracy. The approach achieves the real-time request and has practical value.