Computation of Homography between a Spherical Image and a Perspective Image

While a spherical image has a full field of view with a low resolution with respect to observed scenes, a perspective image has a narrow field of view with a high resolution with respect to observed scenes. Therefore, a spherical image and a perspective image are complementary to each other for the field of view and resolution; a visual system consisting of a spherical image and a perspective image can not only observe the entire surrounding environment, but also focus on a specific object with high resolution. In this paper we investigate the computation of the homography between a spherical image and a perspective image. First, the conventional homography of perspective images is reformulated to the case associating a spherical image to a perspective image. Next, we explain how to carry out feature matching between a full-view image and a perspective image for the computation of the homography. Finally, the experimental results of two applications are given to show the effectiveness of our proposed method.