Research On Video Mosaic Based On Surf

Video mosaic is a technology that stitches the videos, which are taken by some cameras with overlapping perspectives. And it is always used to implement a wide-view and full-view panoramic video. Video mosaic is based on image mosaic, thus, the key technologies of video mosaic mainly include image registration and image fusion. A new method of video mosaic is proposed in this paper. Firstly, two videos captured by two cameras, are treated as sequences of images. Then, feature points of every frame are extracted by SURF. Secondly, the initial feature-point pairs are extracted by the trace of Hessian matrix. And the incorrect feature-point pairs are rejected by RANSAC. Thirdly, the least squares method is used to solve the problem of image transformation matrix. The bilinear interpolation method is used to solve the problem of image interpolation. Finally, the weighted average algorithm of Fade out gradually into is used to remove the parallax in the mosaic image.At the beginning, this paper analyzes feature extraction based on SURF in detail, and proposes a modified version of SURF (M-SURF). It also does lots of simulation in order to verify the superiority of M-SURF. Then, it researches each part of the new video mosaic method which is proposed in this paper. Because the cameras are stationary with invariable the parameter in this paper, transformation matrix of the two images is only counted in the first frame. And other frames can directly use the matrix to mosaic the images no longer feature detection. It greatly reduces the time of image mosaic for each frame. So this method is more close to the requirement of the real-time. According to the design, we implement the video mosaic with VS2008 and OpenCV2.0. And the experimental results show that the proposed video mosaic method is feasible to meet the research requirements. Finally, in order to apply our video mosaic method to the specific intelligent video surveillance in the future, it presents a program of intelligent video surveillance system of autonomous banking network.