Smooth Panoramic Walkthrough for Adjacent Panoramic Viewpoints with Dense Spherical Matching Points

A smooth walkthrough between two adjacent view-points in three-dimensional (3-D) space is the key idea for increasing interactive experiences. It is a relatively easy task when we have the fully 3-D structure of two points. However, most people need a lot of effort to collect 3-D data, and using them efficiently is even harder. In this paper, we propose an algorithm that can create a realistic and natural panoramic walkthrough through two 360-degree panorama images. We estimate the correlated keypoints between the two images through feature matching by using 360-degree images of two points selected by the user. These key points are connected to each other to create faces. By using the spherical coordinate system, the existing key points can be expanded and more detailed faces can be determined. Using the characteristics of the panorama image, we predict the movement of keypoints in the spherical coordinate system, and through this, we predict and create a 360-degree panorama image just as the user moves between the two adjacent images. When moving the space between two points, the users can experience a virtual world through a 360-degree image of the space. With this algorithm, we can access the free view of the rotation and translation of the space by using only the 360° panorama image of the two adjacent points. Hence, we validate our result with the MOS test on various walkthrough situations and our proposed algorithm shows the most preference among users compared to prior algorithms.