Precise positioning of reference image in video frames

In this paper, a new method is proposed to provide a reliable and accurate algorithm to search and precisely find the center of a reference image in a set of video frames. By implementing algorithms such as SIFT and PCA-SIFT, some matched points are obtained. Some of these points are accurate, some of them are not strictly accurate and others are incorrect. As a result, registration process may be inaccurate or mistaken. To remove such errors and inaccurate matched points, we propose a two-stage method. In the first stage, Variance Recursive Test method and in the second stage, an accurate projective model is used. This will lead to detection and removal of the wrong or inaccurate matched points and increase registration accuracy. The precision of this method is about 2-3 pixels in a 600×800 pixel image.