A FAST ALGORITHM FOR AUTOMATED PANORAMIC MOSAICING

Automated panoramic mosaicing method based on cylindrical coordinates is simple and fast, so it is used widely. But there are two problems in this method: the first is that nonlinear least square iterative refining depends on selection of initial value, and often falls in local minima; the second is that not considering the small rotation of camera around optic axis issue results in mis registering between images. A robustness improved fast algorithm is presented for automated panoramic mosaicing. Firstly, using down sampled small images to compute cross correlation, good initial parameters approaching the best solution are obtained and then the parameters are refined by nonlinear least square iterative method. When refining, a parameter of small rotation around image center is added. This algorithm is appropriate for images taken by hand held camera.