Panoramic Image Processing For Virtual Tours

Spherical panoramic images are more and more easily obtained for common users in the past decade.This mainly attributes to the development of image stitching techniques and the maturity of panoramic imaging systems.By providing fields of view far beyond the conventional(planar)images,spherical panoramic images have been widely used in virtual tour systems and lead to the launch of the street view service.To alleviate the visual discontinuity,image blending is used in these systems when transiting between panoramas.However this simple method will make the scene over blurred,and impact the user experience.At the same time,existing systems directly display the raw captured data without editing,which may restrict the applications of panoramic images.To solve these problems,the paper presents panoramic image processing techniques and contains the following contributions.We propose a new fast and robust binary feature for spherical images.Our approach stems from the geodesic grid,a nearly equal area hexagonal grid parametrization of the sphere used in climate modeling.It enables us to directly construct robust features on the hexagonal spherical grid.Extensive experiments show that our method consistently outperforms other existing spherical features in accuracy,efficiency and robustness.To support navigation between cubic panoramas in video viewing mode,we present a panorama interpolation method.Our method extends the matching triangulation interpolation procedure with special considerations of pherical domain.A triangle to triangle homography based warping is developed to achieve physically plausible and visually pleasant interpolation results.The temporal smoothness of the synthesized video sequence is improved by means of compensation transformation.We address a novel problem of cloning a patch of the source panoramic image to the target panoramic image,which is panorama cloning.Considering the sphere geometry constraint embedded in the panoramic images,we develop a coordinate-basedmethod that directly clones in the spherical domain.We discuss in depth two essential issues in panorama cloning,i.e.preserving the orientation of the cloning patch,and handling the large patch cloning which may suffer from discoloration artifacts.As experimental results demonstrate,our method is able to get visually pleasing cloning results and achieve real time cloning performance.